• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

vulngate is an agent-neutral security check tool that scans code repositories for vulnerabilities using SAST, secrets scanning, and dependency auditing. It provides a unified interface for running these checks and offers structured findings with plain English explanations and fix hints. The core of vulngate orchestrates deterministic scanners and normalizes their output, ensuring minimal overhead and cost while providing valuable security insights to developers.

What’s Changed

Follow-up to v1.5.1. That release probed /code-scanning/alerts and treated 404 as “code scanning unavailable” — but a public repo that simply hasn’t uploaded an analysis yet also returns 404 (no analysis found). Result: the very first SARIF upload would be skipped, permanently, on exactly the repos where code scanning works.

Fix: read repository metadata instead of probing for analyses. Upload when the repo is public (code scanning is free there) or when Advanced Security is enabled on a private repo.

Verified against all three real cases: private-without-Advanced-Security → skip cleanly; public → upload.

The v1.5.1 guarantee still holds: a scan that passes never produces a failing build.

• This action is used across all versions by 5 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action is a Yarn 2 plugin designed to simplify dependency upgrades in projects using the Plug’n’Play (PnP) mode. It automates the process of upgrading packages by running yarn up with specific filtering options, such as excluding certain packages or focusing on development or production dependencies. The action also supports specifying range updates for individual packages. The plugin is particularly useful for teams that rely on Yarn 2 and need a straightforward way to manage dependency upgrades without manual intervention.

What’s Changed

• v0.0.5 (fde50fd)
• Update README (ae1cd0e)
• Add more features (0059210)
• Bump-up dependencies (#7) (ee96e21)
• Fix action entry (d872385)
• Upload artifact even release creation was failed (a2983e6)
• Remove unused dependency (6d2524f)
• Fix bump action (382d27e)
• Add publish action (0d0779c)
• Add bumping trigger (0134e22)

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Mutineer is a clean-room mutation-testing tool for Ruby that automates the process of identifying gaps in test coverage by mutating code and checking if tests fail to catch these mutations. It supports various options for customization, such as specifying test files, operators to use, and threshold settings. The action ensures thorough testing by running mutants against test suites and provides clear reports on gaps in test coverage.

What’s Changed

Fixed

• --test-command under version managers — scrub Mutineer’s rbenv/asdf version bins and bundler/gem env from the child so the suite can resolve the app’s Ruby via shims / .ruby-version. Smoke check prints a targeted hint on Bundler::RubyVersionMismatch instead of only blaming DB/migrations. Docs cover a wrapper recipe for stubborn setups (#32).

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 4 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action sets up the Elide runtime, a modern web framework for building APIs and microservices. It automates installation of Elide by default or through various installer methods like archive, shell, apt, etc., allowing users to choose their preferred installation method and customize the setup process. The action also supports caching for faster installations and anonymous error telemetry for better debugging support.

What’s Changed

What’s Changed

• fix: version matching and tool cache keys for build-metadata tags by @melodicore in https://github.com/elide-dev/setup-elide/pull/255
• chore: bump js-yaml by @darvld in https://github.com/elide-dev/setup-elide/pull/256
• chore: update dist by @darvld in https://github.com/elide-dev/setup-elide/pull/257

New Contributors

• @melodicore made their first contribution in https://github.com/elide-dev/setup-elide/pull/255
• @darvld made their first contribution in https://github.com/elide-dev/setup-elide/pull/256

Full Changelog: https://github.com/elide-dev/setup-elide/compare/v4.1.0...v4.2.0

• This action is used across all versions by 8 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Codex Review Action automates the process of reviewing and editing pull requests using OpenAI’s GPT models. It posts inline review comments and a PR-level summary, and can apply focused edits on demand when trusted users comment /codex. The action supports both review and act modes, allowing for minimal or detailed edits based on the specified reasoning effort.

What’s Changed

What’s Changed

• chore: update codex-python to 1.145.0 by @gersmann in https://github.com/gersmann/codex-review-action/pull/25

Full Changelog: https://github.com/gersmann/codex-review-action/compare/v1...v1.8.4

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The action builds, lints, and optionally pushes a Docker image to an AWS ECR repository. It leverages GitHub Actions, OIDC authentication, and IAM roles for secure access to the ECR. The action can handle both push-to-main and release workflows and supports linting and building with optional private npm package dependencies.

What’s Changed

• [skip ci] bump v6.0.4 (dfe428d)
• build(deps): update action-tag-release-build and AWS actions to latest versions (dfa633b)
• [skip ci] bump v6.0.3 (1f89ae4)
• build(deps): bump the actions group with 2 updates (#27) (1af0aef)
• [skip ci] bump v6.0.2 (43d9c12)
• chore: migrate supera.json to 2.x schema (#28) (2a5ab83)
• [skip ci] bump v6.0.1 (35506d7)
• chore: update bats-core action to use version 4.0.0 (200cf9e)
• [skip ci] bump v6.0.0 (08a376d)
• [skip ci] bump v5.0.7 (b34355d)

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of bumping a version number, tagging a commit, creating a release with a CHANGELOG, and optionally updating package.json or Claude Code plugin files. It uses semantic versions to determine the bump type based on merge/HEAD commit messages or defaults to patch. The action can be triggered via a GitHub workflow dispatch event, and it exposes outputs for later use in subsequent steps.

What’s Changed

⚠ BREAKING CHANGES

• feat!: scaffold environment sync command and plan (#48) (23b4ee7177744fbda42d4ebbd2cd43879e0fc1a6)

Features

• feat!: scaffold environment sync command and plan (#48) (23b4ee7177744fbda42d4ebbd2cd43879e0fc1a6)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

NTN-in-a-Box is a self-hostable open-source platform that simulates the conditions of satellite passes, allowing developers to test and build applications under realistic NTN (Non-Terrestrial Network) conditions. It provides a condition engine for shaping real network traffic like a satellite pass and a pluggable module system for building capabilities such as messaging and emergency services.

What’s Changed

Full Changelog: https://github.com/hyavari/ntn-in-a-box/compare/v0.1.4...v0.1.5

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the rotation of Recovery Lock passphrases on Jamf Pro-managed Apple Silicon Mac computers. It uses OAuth client credentials and retrieves device Management IDs from Jamf Pro inventory, with optional scoping via a Smart Computer Group. The action issues the SET_RECOVERY_LOCK MDM command, exposing outputs for downstream workflow steps or reporting. It supports dry-run workflows to validate roles and group scoping before executing MDM commands.

What’s Changed

Jamf Recovery Lock Rotation

This is a version of Inetum-Poland/jamf-recovery-lock-rotation released by GitHub Actions.

Changes since v1.0.0

• 4d11d22 - feat(repo): add release attestation and improve CLI passphrase generation (#6)
• f483467 - docs(action): refine marketplace description (#5)

Full Changelog: v1.0.0…v1.0.2

^{Contributed by @bsojka}

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 38 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action Junie automates the integration of JetBrains’ AI coding agent into GitHub workflows to resolve issues, manage PRs, and provide inline code reviews. It responds to mentions in comments, issues, and PRs, automatically implements fixes and suggestions, and integrates with Jira and YouTrack for project management.

What’s Changed

What’s Changed

• [Junie]: Update Junie CLI Version to 2383.10 in Files by @mashan555 in https://github.com/JetBrains/junie-github-action/pull/184

Full Changelog: https://github.com/JetBrains/junie-github-action/compare/v1...v1.6.1

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Codex Review Gate GitHub Action is a reusable tool designed for repositories that want a required status check to stay pending or failing until Codex review output for the current PR head is clean. It automates the process of requesting and evaluating Codex AI-generated reviews, ensuring reliability and control over the review state machine. The action handles different states of Codex findings, including outdated but unresolved threads, and provides clear disclosure about any generated comments from Codex.

What’s Changed

Fixed

• Accept every exact official clean-result tagline observed during the @v1 consumer rollout.
• Preserve the closed clean grammar: arbitrary prose, punctuation near misses, and finding-formatted content still fail closed.
• Exercise the live Another round soon, please! form through the full state machine, including reasserting success over stale status history without requesting another review.

Release identity

• Action commit: 11d400902175edd773340dc9ec00f8dd421feff7
• Canonical source PR: JoeyTeng/codex-review-gate#21
• Canonical source merge: f8a6b84eb39459dd5aa6c7a20487f444ae387df3

The signed v1.3.2, v1.3, and v1 tags all peel to the same tested action commit.

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Register Jolter Release Action is a GitHub Action designed to register a published plugin release with the Jolter registry. It submits release metadata and the repository-scoped GitHub token, while keeping the plugin artifacts hosted in the GitHub Release. The action automates the process of registering releases for plugins, handling semantic versioning, and ensuring safe re-runs when versions are already registered.

What’s Changed

Breaking change

With this release we are upgrading the typescript version from v6 to v7 for building the project. This does not affect how the action works, it only affects the development workflow of this action.

What’s Changed

• deps: bump actions/setup-node from 6 to 7 in the github-actions group by @dependabot[bot] in https://github.com/jolterjs/register-release-action/pull/6
• deps: bump typescript from 6.0.3 to 7.0.2 by @dependabot[bot] in https://github.com/jolterjs/register-release-action/pull/5

Full Changelog: https://github.com/jolterjs/register-release-action/compare/v1...v1.1

• This action is used across all versions by 3,362 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The datamodel-code-generator GitHub Action generates Python data models from various schema definitions including OpenAPI 3, AsyncAPI, JSON Schema, Apache Avro, XML Schema, Protocol Buffers/gRPC, GraphQL, and MCP tool schemas. It can also convert existing Python types like Pydantic, dataclass, and TypedDict into different output types such as Pydantic v2, Pydantic v2 dataclass, dataclasses, TypedDict, or msgspec. The action handles complex schemas with features like $ref, allOf, oneOf, anyOf, enums, and nested types to produce type-safe, validated code ready for IDEs and type checkers.

What’s Changed

Breaking Changes

Default Behavior Changes

• New warning emitted by default for unresolved local $ref pointers - By default, an unresolved local $ref JSON pointer now emits a new DanglingRefWarning and generates a fallback Any model. Previously such refs resolved to a silent fallback, and some cases (e.g., out-of-range JSON pointer array indices) raised a hard error. Schemas that previously generated cleanly can now produce warnings, which breaks workflows that treat warnings as errors (e.g., python -W error). Use --strict-refs to fail on unresolved pointers instead, or --disable-warnings to silence them (#3639)

Error Handling Changes

• $ref to a non-dict file now raises InvalidFileFormatError instead of TypeError - Resolving a $ref to a JSON/YAML file whose top-level value is not a mapping (e.g., a list) now raises datamodel_code_generator.InvalidFileFormatError (a subclass of Error/Exception) rather than the built-in TypeError. Programmatic callers that catch TypeError around generate() must catch InvalidFileFormatError (or Error) instead (#3639)
• Malformed-input diagnostics reworded - Malformed inputs that previously surfaced an uncaught traceback now emit a concise message such as Invalid file format for <type> at <source>: ..., and the missing-file message changed from File not found to File not found: <path>. Tooling that matches on the exact previous strings needs updating (#3639)
• Generation now aborts when output paths would overwrite inputs or collide - A new path-conflict validation runs before generation in both the generate() Python API and the CLI. Invocations that previously succeeded now raise an Error (CLI exits with an error code) in these cases: the output path resolves to an existing input file, the --emit-model-metadata path resolves to an existing input file, or the output and model-metadata paths resolve to the same file. Symlinks and hardlinks are resolved before the comparison. Workflows that intentionally wrote output over an input path, or that pointed the model-metadata artifact at the same path as the output, will now fail with one of:

Output path must not overwrite an input path: <path>
Model metadata path must not overwrite an input path: <path>
Output and model metadata paths must be different: <path>

(#3647)

What’s Changed

• Update CHANGELOG for 0.70.0 by @dcg-generated-docs[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3619
• Decouple RootModel dependency ordering by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3620
• Improve README and docs homepage links by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3624
• Update release benchmark data by @dcg-generated-docs[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3625
• Move discriminator policies to output models by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3621
• Invalidate generation facts for every list mutation by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3622
• Document package manager installation options by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3629
• Move dict-key dependency policy to output models by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3626
• Decouple parser model behavior helpers by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3632
• Share model constraints across output backends by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3630
• Move Pydantic type rendering to backend by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3627
• Move dataclass ordering policies to output models by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3628
• Move template reference metadata to output models by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3631
• Declare output model construction capabilities by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3623
• Fix payload runtime validation exclusions by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3635
• Cover output backends across input formats by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3636
• Track large schema memory by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3637
• Reuse remote schema connections by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3638
• Improve malformed input diagnostics by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3639
• Bump setuptools from 82.0.1 to 83.0.0 by @dependabot[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3633
• Update project usage examples by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3641
• Fix vLLM usage link by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3642
• Refresh custom template paths by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3640
• List project maintainers by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3643
• Preserve Unicode line separators by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3644
• Fix TypedDict unique item sets by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3645
• Defer msgspec forward references by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3646
• Validate generation path conflicts by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3647
• Enforce dynamic model cache size by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3648
• Filter imported dynamic model classes by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3649
• Defer nested model default factories by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3650
• Isolate generation state by @koxudaxi in https://github.com/koxudaxi/datamodel-code-generator/pull/3652

Full Changelog: https://github.com/koxudaxi/datamodel-code-generator/compare/0.70.0...0.71.0

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action exchanges the workflow’s identity token for a GitHub token from the OctoSTS service, using a configured trust policy. It automates tasks such as federating tokens and configuring Git authentication based on the provided trust policy. The key capabilities include setting up federated authentication and handling default configurations.

What’s Changed

What’s Changed

• fix: Restore host-wide configure-git credential by @jsonbailey in https://github.com/launchdarkly/octosts-action/pull/20

New Contributors

• @jsonbailey made their first contribution in https://github.com/launchdarkly/octosts-action/pull/20

Full Changelog: https://github.com/launchdarkly/octosts-action/compare/v1...v1.4.1

• This action is used across all versions by 4 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary:

The setup-fortran GitHub Action automates the installation and configuration of Fortran compilers across Linux, macOS, and Windows runners. It supports various compilers such as GNU Fortran (gfortran), Intel Fortran (ifort), and NVIDIA Fortran (nvfortran). The action allows users to specify which compiler they want to install and its version, and provides options for MSYS2 subsystem configuration and disk space cleanup for certain compilers like NVFortran.

What’s Changed

• Add support for lfortran 0.64.0.
• Robustness improvements.

• This action is used across all versions by ? repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The latexbuild GitHub Action automates the compilation of LaTeX documents by using a Docker container with a full TeX Live distribution. It simplifies the process of compiling .tex files locally without requiring any installation, making it easy to automate build processes or live preview PDFs as you edit your source files.

What’s Changed

Full Changelog: https://github.com/MrzAhmadi/latexbuild/compare/v0.0.5...v0.0.6

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Test-Guard automates the task of ensuring that code changes are adequately covered by tests. It uses diff coverage, heuristic matching across multiple languages, and AI-powered evaluation to ensure that every modified file has corresponding tests or is trivially untestable. The action provides a single pass/fail verdict based on these evaluations, helping developers quickly identify whether their PR has sufficient testing.

What’s Changed

Patch release. If you use a jest/istanbul clover.xml, v1.10.0 did not deliver its own fix — upgrade to this.

The bug

v1.10.0 taught Test-Guard to read the coverage report directly, so a file with no executable changed lines (type declarations, interface members, doc comments) would pass instead of failing with “not in coverage report.” For jest reports, that check still failed.

Clover reporters disagree about which attribute holds the qualified path:

<!-- jest / istanbul -->
<file name="types.ts" path="/home/runner/work/app/app/src/bruno/types.ts">

<!-- PHPUnit -->
<file name="/var/www/app/lib/AuthCookie.php">

The presence check read name=, which is right for PHPUnit but yields a bare basename under jest. is_in_report() then refused to match it — deliberately, since a bare basename would otherwise match a same-named file in an unrelated directory. Net effect: a file sitting in the report at 100% was still reported as un-instrumented.

The fix

The Clover presence extractor now reads both name= and path= from every <file> and lets is_in_report() pick whichever is qualified. Path-prefix detection and Clover rewriting are untouched, so nothing else changes.

Diagnosed and verified against a real failing CI artifact, not a synthetic fixture: is_in_report("src/bruno/types.ts") returns True, Layer 1 returns PASS and short-circuits, and Layer 3’s shortcut returns SKIP where it previously returned FAIL. A path genuinely absent from the report still fails, so real instrumentation gaps are still caught.

Regression tests now pin both reporter shapes.

Upgrading

No configuration changes. Users on @v1 get this automatically on the next run. If you added an exclude entry to work around a declaration-only file, you can drop it.

Full changelog: https://github.com/Ostico/test-guard/compare/v1.10.0...v1.10.1

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action converts an uploaded plugin zip artifact into a publicly accessible WordPress Playground link, allowing developers to easily test and deploy their plugins. It automatically handles private-repo artifacts by failing the action if the repository is private. The action supports various optional configurations such as activating themes and plugins upon installation, and provides options to post the link as a sticky comment on pull requests or in GitHub comments.

What’s Changed

First public release.

WordPress Playground Link turns a plugin zip you’ve already uploaded as a GitHub Actions artifact into a one-click WordPress Playground link — install, activate, and land straight on the page you want reviewers to see, no manual setup.

What it does

• Resolves your artifact to a public URL via nightly.link and builds a Playground blueprint around it
• Installs and (optionally) activates the plugin, with configurable landing page, WP/PHP version, and multisite
• Can install extra plugins or a theme alongside it
• Optionally posts/updates a sticky PR comment with the generated link
• Verifies the artifact actually exists before building a link, if you pass a token — fails loudly in CI instead of shipping a dead link

Three ways to trigger it

Manual (workflow_dispatch), automatic on every PR push, or label-triggered on demand — all three have complete, ready-to-copy example workflows in the README.

Requirements

Your repository and artifact must be public — nightly.link can’t proxy private-repo artifacts.

Full Changelog: https://github.com/pattonwebz/wordpress-playground-action/commits/v0.1.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Postman Onboarding: Workspace Bootstrap action automates the creation of a Postman workspace from an OpenAPI specification. It generates baseline, smoke, and contract collections with executable tests, covering various protocols including gRPC, SOAP, GraphQL, AsyncAPI, and MCP lanes. The action handles spec imports securely through access tokens and supports both public HTTPS URLs and local file paths for the OpenAPI document.

What’s Changed

What’s Changed

• Omit converter-only options from Spec Hub links by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/143
• Release v2.10.20 by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/144

Full Changelog: https://github.com/postman-cs/postman-bootstrap-action/compare/v2...v2.10.20

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

ReleasePilot Guard is a GitHub Action designed to assess the release impact of pull requests in Expo projects. It evaluates changes and provides a safety classification (OTA_safe, native_build_required, etc.) based on Git history and project configurations. This helps ensure that only safe updates are merged, reducing potential risks associated with dynamic or unverified code changes.

What’s Changed

First public release

ReleasePilot Guard is a conservative release-impact checker for Expo applications.

It analyzes a pull-request Git range and classifies the safest release route as:

• OTA safe
• Native build required
• Backend or configuration dependency
• Store metadata only
• No app release required
• Manual review

Included

• Reusable composite GitHub Action
• Real Git-range analysis
• Nested Expo monorepo support
• Dependency and lockfile detection
• Static Expo and EAS configuration comparison
• Machine-readable evidence report
• Conservative handling of unknown changes

Validation

Tested against 25 pinned commit ranges from five public Expo repositories:

• 22/25 automatically classified
• 6/6 reviewed OTA-safe classifications correct
• Zero unsafe changes labeled OTA-safe

ReleasePilot is advisory by default and does not deploy applications or execute untrusted pull-request code.

• This action is used across all versions by 2 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Outrider is a GitHub Action that automates the process of evaluating new methods against an existing codebase. It helps teams validate and compare new implementations by scheduling or dispatching evaluations on demand, using Anthropic Opus or z.ai GLM-5.2 as model backends for cost-effective evaluation. The action provides draft PRs with self-reviews, issues when preflight, validators, or self-review routes the paper to discussion, a branch-only mode, and ensures no duplicate work by not re-recommending papers once referenced.

What’s Changed

Adds first-class support for REVIEW.md — a per-repo review-conventions file that Outrider threads into the drafter, fidelity audit, and convention pass on dispatch.

Changes

• New REVIEW.md at repo root documenting this repo’s own review conventions (#107)
• Reader + orientation-block threading: _orient_review_conventions() reads REVIEW.md (root wins over .github/) and inserts it as the highest-precedence signal in the orientation block, ahead of contributor guides (#108)
• Fidelity audit + convention patch prompts cross-check against REVIEW.md’s scope-stance and anti-pattern sections when present (#108)
• docs/REVIEW-template.md — starter template maintainers can copy into their repos (#108)
• 8 tests covering discovery, precedence, empty-file, truncation, orientation-block inclusion/omission (#108)

Silent info-level fallback when REVIEW.md is absent; no failure. Existing repos without a REVIEW.md see unchanged behavior.

• This action is used across all versions by 5 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Bernstein is a deterministic orchestrator that automates CLI coding tasks using various adapters. It provides reproducible runs by scheduling them in plain Python and ensures checkable results through an always-on lineage spine and replay journal. The action supports multiple agents, including Claude Code, Codex, Gemini CLI, and 40+ more, with options for local and air-gap installations and auditing receipts offline.

What’s Changed

A feature release. This wave widens the verifiability surface: the run journal, lineage spine, and HMAC audit chain gain new first-class receipts across identity and delegation, outbound payments, non-coding artifacts, evaluation, and cross-agent interop, and a hardening pass runs through the event, admission, MCP-audit, cost, and process-reap surfaces. No existing public API changes; upgrade in place, but read the behaviour-change note below first. The exhaustive index is docs/reference/FEATURE_MATRIX.md.

Behaviour changes (read before upgrading)

• A halted mission phase no longer halts the whole mission, and is no longer reported as the active phase (#2683). Phases already run under isolated budget envelopes at dispatch, so exhausting one phase’s envelope never gated a sibling’s spend. The status projection had not caught up: it reported a halted phase as the active phase and folded any halt into an overall halted mission, so one exhausted envelope masked a runnable sibling. The projection now skips halted (and passed) phases when selecting the active phase, and reports the mission halted only when no phase remains runnable. A halted phase beside a runnable one now leaves the mission active with the runnable phase reported active.

  The active phase and the overall mission state are both hashed into mission_status_hash, so this correction deliberately moves that hash for any ledger carrying a halt, and for every ledger it moves the schema-version field alone. MISSION_STATUS_SCHEMA_VERSION is therefore bumped from 1 to 2. That field is how a verifier tells “folded under newer projection rules” apart from “tampered with”: an archived ledger stays reproducible because project_mission (and project_mission_from_ledger) accept an explicit schema_version, and projecting an archived ledger under schema_version=1 reproduces its original status hash byte for byte. The pinned v1 goldens assert exactly that reproduction.

  Halt semantics are unchanged and stay chain-anchored: a halted phase still seals its halt receipt onto the work ledger and mirrors it onto the HMAC audit chain. Isolation means a sibling phase continues, not that a halt is weakened. The mission daily-progress digest embeds mission_status_hash, so its digest_hash and receipt_id move with the projection even though the digest document schema itself is unchanged.

Identity and delegation

• Attenuated delegation capability tokens: each hop is a detached-JWS token over a JCS-canonical body whose caveats can only narrow (permissions, task-id allowlist, path prefixes, expiry, uses, depth), verifiable offline from the token bytes with the registry unavailable, and minted as a chain-anchored event (#2901).
• The install-identity signature on outbound agent-facing requests now binds the request body (an RFC 9530 Content-Digest folded into the RFC 9421 covered set), and the per-hop delegation ledger serialises its read-modify-write under a cross-process lock so concurrent writers can no longer fork the chain (#2935).
• A2A server surface: a signed capability card served at both well-known paths, message/send + tasks/get over JSON-RPC with SSE, auth declared in the card, and every inbound task returning a chain-anchored lineage receipt an external caller can verify offline; clients without streaming or artifacts still work through polling (#2911).
• AGNTCY ADS publication: bernstein a2a publish --surface agntcy-ads emits an OASF capability descriptor that is a deterministic projection of the signed card, carrying a provenance attestation signed by a trust root distinct from the card key; republishing an unchanged node is byte-identical (#2943).

Payments and cost governance

• Signed payment mandates with chain-anchored receipts: a SpendMandate is an Ed25519-signed, content-addressed authorization bounding amount, currency, recipient, and expiry in a human_present or delegated presence mode; every attempt is a TransactionReceipt (a refusal is a first-class receipt with a closed reason), amounts are string-encoded integer nano-units with no float in a signed payload, and cumulative spend is aggregated under a lock so concurrent workers sharing a mandate cannot exceed the cap (#2902). The mandate and consent-receipt surfaces were then hardened (#2936).
• x402 settlement hook: metered MCP gateway calls settle through the mandate framework, binding the signed intent, the authorized calls, and the settlement reference into one chain-anchored consent receipt (#2937).

Non-coding artifacts and evaluation

• Typed artifact contract (slice 1): an ArtifactKind / ArtifactSpec threaded onto Task (defaulting to code_diff), widened lineage artefact kinds, per-kind canonical serialisers over one shared core, three new completion signals (schema_valid / criteria_match / hash_stable), and bernstein artifact verify — so a report, dataset, or action log is a signed, content-addressed lineage record rather than a blob (#2903).
• Figure grounding for report artifacts: every material number in a report must trace to an anchored source (attachment, artifact, or query receipt) or the task does not complete; the figures sidecar is inside the artifact’s own content hash, and bernstein artifact verify renders per-figure provenance (#2912).
• Content-addressed query receipts: a read-only SQL result set an agent consumed is canonicalised and its hash bound into a signed lineage entry; bernstein datasource verify proves it offline and --re-execute reports MATCH or DRIFT, and DML/DDL is refused (#2909).
• Deterministic replay-debugging surface: bernstein replay debug walks the Merkle step chain to name the exact diverging step and field, emits a byte-identical two-run path-diff, forks from a step, and produces an offline-verifiable debug receipt (#2900).

Cluster and lifecycle

• Leaderless MESH claim journal (phases 1-2): a signed, append-only, hash-chained claim journal with a pure project_claims fold that two nodes reduce to byte-identical state, and a deterministic lowest-entry-hash supersede rule; the SQLite ledger becomes a projection of the journal on the opt-in path while STAR behaviour is unchanged (#2907).
• Mission phase isolation is described in the behaviour-change note above (#2683).

Integrations

• Generic OData v4 system-of-record integration: a watermark/delta trigger source with a deterministic, chain-anchored poll cursor, and receipt-anchored write-backs (GET-before-PATCH with If-Match, typed 412/428 conflicts, draft-activate) where every write-back is a signed audit event verifiable by bernstein audit verify (#2908).
• Self-hosted OpenAI-compatible endpoint certification: bernstein endpoints certify --base-url produces the existing signed certification record for an arbitrary endpoint, with a docs page covering the certify → verify loop and the exercised endpoint families (#2913).

Hardening and security

• Bounded hardening passes through the event feed, triggers, webhooks, and receipts (#2890); the pool/lease admission engine, ledger, tags, and verifier (#2891); MCP stateless/gateway/client audit ordering (#2892); provider-state mutation capture (#2893); cost-scheduling receipts and the price table (#2894); and process reap, platform compat, and worktree isolation (#2895).
• bernstein audit verify is now total over an undecodable chain segment: a non-UTF-8 segment is reported as a named verification failure at the exact byte offset instead of raising out of verify() (#2880).
• bernstein lineage verify fails closed with a distinct exit code when the audit key is missing, instead of misreporting a setup problem as tamper (#2882).
• The SLA-receipt log sink escapes caller-supplied paths so a CR/LF path can no longer forge log lines (#2881); adapter capability-profile selection is recorded at spawn so replay detects drift (#2883).
• The CI weekly-digest and skills-publish workflows were hardened, along with skills packaging (#2945, #2946).

Observability

• The OpenTelemetry GenAI-semconv gaps in the signed OTLP span projection were closed (#2896).

CLI and documentation

• Operator-surface follow-ups from the 3.8.2 sweep: an honest dashboard capacity denominator, stop --force process-group reaping, a live-scan ps cross-check, and a bernstein cluster status node view (#2897).
• The README now front-loads the four differentiators as a scannable at-a-glance ahead of the full capability list (#2938); the feature matrix was refreshed for currency across the verifiability, identity, payments, and provenance surfaces shipped since the last pass (#2941); the agent-doc mirrors gained currency guards so the generated views cannot silently drift (#2906).
• A decision record measures the cost of enabling the HMAC audit chain by default and defines the migration path, without flipping the default (#2905).

Thanks

Thanks to @Maqbool61 for documenting and helping certify the self-hosted OpenAI-compatible endpoint path (#2913).

pip install --upgrade bernstein==3.9.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

GitGalaxy automates security and architecture assessments for repositories, offering a deterministic audit tool that can scan multiple languages and detect risk exposures in full repos, providing actionable fixes to lower those risks. It uses a custom regex/lexical structural-analysis engine to identify code patterns relevant to security and architecture, generating normalized 0–100 risk scores and exporting the results in various formats for further analysis.

What’s Changed

GitGalaxy v2.4.5: The Correctness Release

This is a dense one. Where v2.4.0 was about getting GitGalaxy’s findings in front of more people (SARIF everywhere, every CI platform), v2.4.5 is about making sure the findings underneath are actually true. It closes two long-running epics, fixes a critical bug that had silently disabled a core engine for every one of GitGalaxy’s 58 supported languages, patches a disclosed CVE, and knocks out ~20 smaller data-integrity fixes along the way.

Critical Fix: The Comment/Code Separation Engine Was Broken for Every Language (#386, #390)

What we found: My previous workflow resulted in Gemini based bug patterns of consumer producer mismatches of keys between files. Explicitly revealed with prism.py, the module responsible for stripping comments from source before analysis, was looking up its per-language delimiter rules under config keys (mechanical_families, recursive_c_style, column_sensitive, single_line_only, c_style_comment) that no producer in the codebase had ever written. The real config used a different taxonomy entirely. Every lookup silently fell through to defaults, meaning comment stripping had been non-functional — or working by pure coincidence — across all 58 supported languages since the feature was introduced.

Why nothing caught it: test_prism.py’s test fixtures were hand-built mocks that mirrored the wrong key names, so they were internally consistent with the bug and passed cleanly for hundreds of runs. Real production config and test config had quietly diverged.

What we did: Renamed the lookups to match the real taxonomy (lexical_families, recursive_block, positional_anchored, line_exclusive, standard_block), replaced the mocks with tests that exercise the actual LANGUAGE_DEFINITIONS/LEXICAL_FAMILY_HEURISTICS config, and — in a follow-up (#390) found while double-checking the first fix — caught and fixed one more missed reference in _strip_nested_comments() that had been masked by a fallback value that coincidentally matched the correct data. The new regression test uses non-default delimiters specifically so that kind of coincidence can’t hide a bug again.

Known limitations, documented and deferred on purpose: a 9-delimiter gap in standard_block affects 5 non-C-style languages, and golden_master.json was not regenerated as part of this fix. Both are scoped as explicit follow-ups rather than folded in here.

New: The Dead Key Auditor, and the Closure of Epic #325

Epic #325 tracked a recurring failure pattern across the codebase: a consumer reads data["some_key"], but no producer anywhere ever writes "some_key" — a silent, type-checker-invisible class of bug.

• Built tests/dead_key_audit.py: an AST-based static analyzer that walks the codebase, cross-references every dict-key read against every dict-key write (including f-string prefix patterns like f"sec_{key}"), and reports the gaps.
• Wired it into CI (dead-key-audit.yml) as a baseline-gated required check — new dead keys fail the build, but the tool doesn’t demand the whole codebase be clean on day one.
• Ran it, and fixed all 15 findings it surfaced, filed and closed individually (#364–#378):
  • Every recorder read "AI Threat Score"; the producer wrote "AI Threat Confidence" (#364)
  • AIAppSecSensor’s Over-Permissioned Agent rule was reading a dead ai_tools signal instead of llm_orchestrator (#365)
  • Four dead security-flag SQLite columns in record_keeper.py (#366–#369)
  • repo_z_score and typosquat_hits were computed but never backfilled into the summary, in both the main and Delta Mode pipelines (#370, #371, #376)
  • max_big_o was written onto the graph node but never copied back onto the file dict (#372)
  • has_tests was a dead read; wired to real test_coverage_map (#373)
  • aperture_reason vs. reason key mismatch across signal_processor.py/audit_recorder.py (#374)
  • Missing TYPOSQUAT_WHITELIST in APERTURE_CONFIG (#375)
  • Missing cross-referenced disqualifiers for matlab and objective-c language definitions (#377)
  • The mechanical_families investigation above, which escalated into #386 (#378)

The baseline file is now empty — every finding the tool surfaced this cycle has been resolved.

Epic #102 / #348: Spatial Correlation, Phase 2–3

Continued work on satellite-aware, post-hoc spatial correlation of security signals:

• Persisted a satellite-aware spatial ledger for post-hoc correlation across scans (#348)
• Scoped dampener correlation to real function boundaries instead of naive proximity (#346, phase 1)
• Implemented the Spatial Radar for DB injection funnels (#105)
• Spatially verified Metaprogramming Hallucination Risk in the Dev Agent Firewall (#106)
• Scoped the Exfiltration Distance Check to real function boundaries (#102)
• Fixed detector.py’s dead RCE-taint corroboration and a sec_ merge overwrite bug that had been silently dropping signals (#344)

Config Resolver: Phase 0–1

Started consolidating GitGalaxy’s config precedence rules (CLI flags, YAML, env, defaults) into a single, testable merge point:

• Added config_resolver.py with a single ResolvedConfig merge point (Phase 1)
• Migrated galaxyscope.py’s main() and 7 direct gitgalaxy_config.py importers off scattered module-level constants

Security

• Patched a disclosed CVE (GHSA-9xwg-3r6f-jcx2 / CVE-2026-61632, path traversal in pymdown-extensions’s b64 extension) by bumping to 11.0.1
• Fixed churn risk silently scoring zero for docs and critical-secret-leak files (#245)
• Fixed a CLI/YAML sentinel-collision bug in the pipeline-flag interceptor that could cause flags to be silently ignored (#247)
• Made Aperture’s file-size gate tiered and intent-aware instead of a single blunt cutoff (#343)

Precision & Data-Integrity Fixes

A long tail of quieter correctness fixes, several with real blast radius:

• IGNORED_DIRECTORIES case-mismatch was silently disabling 9 ignore rules (#306); GuideStar wasn’t even pointed at the shared set (#319)
• Case-insensitive matching for shadow-API file extensions (#320) and SBOM malware-scan candidates (#321)
• Confidence/exposure scores that could exceed their intended 0–100 or 0–1 scale now clamp correctly (#310, #318, #331)
• Removed several dead reads that had been silently no-op’d for a while: func_empty/keyword_debt in tech-debt scoring (#308), ai_tools/ai_memory/ai_logic_loop from SIGNAL_SCHEMA (#323)
• Fixed zero_dependency_mode under-reporting during incremental scans (#307)
• AIAppSecSensor was reading its 8 inputs from a phantom telemetry namespace that nothing wrote to (#326)
• duplicate_logic count wasn’t being aggregated into equations the way orphaned_logic was (#327)
• Wired opt-in network-centrality risk weighting into the firewall (#304), and the supply chain firewall into SignalProcessor’s risk vector (#302)
• Fixed a file-miscounting bug for files sharing the same name (#261/#273)
• Fixed docstring harvesting running past a stand-alone closing delimiter (#279)

Type Safety

Closed out a focused mypy-hardening pass: fixed indexing and type-spoofing errors across state_rehydrator.py, cobol_jcl_auditor.py, llm_recorder.py, galaxyscope.py, and FunctionNode’s TypedDict schema (#293–#301), plus a mass annotation cleanup.

Supply Chain / SBOM

• Unified manifest-filename registries, fixing mixed-ecosystem SBOM starvation (#285)
• Added a dependency audit cache and entry-point priority audit ordering (#283, #284)
• Consolidated manifest parsing into manifest_parser.py

Docs & CI

• Corrected docs/CI-template accuracy repo-wide (“Honest READMEs”) and added a GitHub Actions reporting stage (#264)
• Fixed the LLM architectural brief bot’s identity and its workflow (it now opens a PR via a forked create-pull-request action instead of pushing directly to protected main)
• Pinned galaxyscope --output to a file, not a folder, to match downstream SARIF/SBOM/LLM paths (#265)
• Bumped Muninn (our CI security scanner) to v0.3.5 and then v0.3.6

Full changelog: v2.4.0...v2.4.5

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action downloads and runs Ternary Bonsai 27B locally on a GitHub Actions runner, allowing inference without sending prompts or responses to Hugging Face. It caches the model and runtime for subsequent runs, supports local CPU inference, and provides options for prompting through text or file inputs.

What’s Changed

Initial local inference release

• Download and verify the pinned 7.17 GB Ternary Bonsai 27B Q2_0 GGUF on the runner.
• Cache the model and pinned PrismML llama.cpp runtime.
• Run inference through a loopback-only local llama-server.
• Return the final answer, reasoning, timings, model path, and response file.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

RustScript is a Rust runtime that interprets and executes Rust scripts without compiling them to machine code. It provides a way to run small Rust programs directly from the command line or as part of CI/CD pipelines. This allows developers to quickly test snippets or automate tasks without waiting for a full compile.

What’s Changed

Full Changelog: https://github.com/VladasZ/rustscript/compare/v0.2...v0.2.12

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This action automates the testing of Claude Cowork skills in a headless, scriptable, and CI-friendly environment. It reproduces the observable runtime contract closely enough to test skills across various scenarios without relying on the locked Desktop app. The main purpose is to reduce the cost and complexity of skill development by allowing for automated testing with limited or no access to the Desktop app itself.

What’s Changed

Added

• schema/critique-report.json — a descriptive, test-pinned schema for critique’s JSON report / critique-report.json artifact, so automation consumers (budget pacers gating on costUsd.complete, harvesters) parse field names/shapes from a schema instead of prose. Deliberately not a SPEC §12-frozen surface (unlike doctor.json) — critique is EXPERIMENTAL and additive field changes are expected; the schema says so in its own description, and a two-way sync test pins it against the actual report builder on every branch (findings / infraFailure / evaluatorError).
• gradedSkill in the critique report (text header + JSON): the resolved skills/<name> the packager graded under --skill/auto-selection. Load-bearing for multi-skill plugins: gradedSkillHash keys the whole mounted plugin, so pairing by hash alone cross-pairs critiques of different skills — pair by (gradedSkillHash, gradedSkill). Docs updated accordingly.

Docs

• Truncation→DROPPED mechanics: a finding whose evidence quotes SKILL.md text past the packaging cap fails citation-resolution and lands in DROPPED (the check runs against the cut copy) — documented next to skillMdTruncated so a back-half DROPPED skew on an oversized skill has its cause named.
• findingFingerprint direction-of-inference: high-precision, LOW-RECALL — a match proves reproduction; a mismatch does NOT prove non-reproduction (the same finding reworded fingerprints differently). The Reproduction section says so before anyone concludes “didn’t reproduce”.
• Evaluator cost share: the two evaluator passes were ~3/4 of a measured e2e total — the calibrate-then---evaluator-model strategy is now in the cost section and critique --help, with the armor-verification-is-default-evaluator-only caveat.
• SPEC §12 now names the critique report explicitly under NOT covered: schema/critique-report.json is descriptive (parse against it, not prose) but not the compatibility contract while critique is EXPERIMENTAL — its surface-baseline presence is for change visibility, and it is the promotion candidate on the doctor.json template once critique stabilizes. README, llms.txt, and the shipped skill point at the schema and the (gradedSkillHash, gradedSkill) pairing rule.

Fixed

• critique’s “Attached inputs” evidence no longer reports (none) when mounts.json is corrupt. listAttachedInputs derived connected-folder names from loadVmPathContext, which returns null for BOTH an absent mounts.json (legitimately no mounts) and a present-but-unparseable one (the folder map is UNKNOWN). Uploads already distinguished these (the ENOENT-vs-read-fault split), but folders — which have no fixed-layout fallback — silently collapsed to [], so a corrupt mounts.json rendered (none), telling the evaluator “the agent correctly saw no connected folder” when the truth was unknown. It now surfaces a corrupt mounts.json as an explicit UNKNOWN note, completing the same confabulation-vs-correct guard the uploads path already applies.
• diff no longer treats two tool inputs that differ only past the ~2000-char cap as the same call. canonicalizeInput truncated a tool input’s canonical JSON to a 2000-char cap and used that truncated string as the tool-sequence equality key, so two Write/Bash calls sharing a long identical prefix but differing only in the dropped tail compared as op: "same" in diffToolSequence — a false “no change” that could flip the advisory diff exit code to 0. A truncated key now folds in a #<len>·<sha16> hash of the full canonical string, so the key depends on the entire content while the visible prefix stays readable in the hunk; both diff sides canonicalize identically, so the comparison stays consistent.
• COWORK_VM_GATEWAY is now validated as a canonical IPv4 literal. The L2-microVM gateway override was interpolated verbatim into a root-run guest iptables -A OUTPUT -d <gateway> command (via sh -c), so a malformed or hostile value could inject shell syntax into privileged provisioning — or, more mundanely, leave the firewall in an unknown state. vmGatewayIp() now rejects anything that isn’t a canonical IPv4 literal (digits-and-dots only, octets 0–255, no leading zeros), failing loud instead of reaching the shell. Operator-set env var, so this is defense-in-depth; no valid gateway value is affected.
• The agent.stderr.log sink is now flushed before the teardown secret-scrub reads it. The stderr sink was piped fire-and-forget and never awaited, so bytes still buffered when scrubRawRunLogs read the file could land raw afterwards — a persisted-secret leak in a narrow teardown window. LiveAgentSession now pipes it with { end: false } and ends+awaits it in the same session-teardown drain that already flushes events.jsonl / control-out.jsonl, so the session generator resolves only after the sink is fully flushed — the scrub always sees the complete log.
• critique no longer prints raw host paths in its report or diagnostics. The text report’s run dir: line, the inspect <dir> hints, the write-failure diagnostics, and the echoed skill-folder path all printed absolute $HOME-rooted paths, so a shared report or screenshot leaked the username + filesystem layout (it landed in a video frame). critique was the one rendering path in the CLI that never called tildeify, while skill/run scrub unconditionally. Every human-facing path is now collapsed to ~; the JSON report and persisted-artifact paths stay raw (machine data a consumer feeds back to a tool). The --demo rejection is unchanged — it was never the fix — but its reason now notes the report already collapses paths.
• critique fails fast on a missing or non-directory skill folder. A typo’d/absent positional folder previously minted a session and spawned the task turn before infra-failing (exit 2), leaving a stray run dir behind. resolveCritiquedSkillDir now existsSync/isDirectory-checks the folder up front — before any session is minted or spawned — so a bad path exits 2 immediately with nothing left on disk. A present-but-SKILL.md-less folder still defers to the packager’s degraded flow, unchanged.
• critique’s report header and stderr diagnostics now say critique: (were skill-critique:). A leftover label from the scripts/skill-critique.ts instrument; the invoked command is critique. Cosmetic, no schema change.

What’s Changed

• release: 1.9.0 by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/74

Full Changelog: https://github.com/yaniv-golan/cowork-harness/compare/v1...v1.9.0

• This action is used across all versions by 303 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action runs specified cargo commands on a Rust project. It automates the execution of common tasks such as building and testing, and provides transparent support for cross-compilation through optional tools like cross or cargo-hack.

What’s Changed

Patch release with updates to vulnerable dependencies.

What’s Changed

• chore(deps): bump fast-xml-parser from 5.9.3 to 5.10.1 in the npm_and_yarn group across 1 directory by @dependabot[bot] in https://github.com/clechasseur/rs-cargo/pull/427
• fix: update @clechasseur/rs-actions-core to 8.0.3, run npm update to get vulnerability fixes by @clechasseur in https://github.com/clechasseur/rs-cargo/pull/429

Full Changelog: https://github.com/clechasseur/rs-cargo/compare/v5.0.6...v5.0.7

• This action is used across all versions by 35 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action compares versions in project manifests and checks they are higher than the latest published tag, supporting npm, Cargo, and Poetry package managers. It uses GitHub API tokens to access tags and supports filtering tags with a regular expression. If the local version matches the latest published tag, it can either fail or return a boolean indicating if it’s a new version.

What’s Changed

Changes

chore(deps): update npm - all minor and patch updates

• This action is used across all versions by 2 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The moodle-github-sync-1 GitHub Action automates the synchronization of Moodle themes and plugins with their corresponding files on GitHub. It solves the problem of maintaining and updating Moodle resources by allowing developers to push changes from local repositories directly to the remote repository, ensuring consistency across different development environments and platforms.

What’s Changed

• Update action.yml (dac9fb3)
• fix (4dfa764)
• fix (e87bf87)
• fix (e724dd2)
• fix (d7578cc)
• fix (2fd3a3b)
• fix (ed97de6)
• Update README.md (0aa4988)
• fix (c97812f)
• fix (e3035b7)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action builds a container image and produces an SBOM, refuses to push it if it fails a vulnerability threshold, and signs the result with keyless attestation using pinned commit hashes, ensuring security by maintaining immutable evidence. It addresses trust issues in the supply chain by preventing malicious pushes through mutable tags.

What’s Changed

Shortens the action.yml description below the GitHub Marketplace limit of 125 characters. No behavioural change from v0.1.0.

Released as a new version rather than by re-pointing the v0.1.0 tag. Moving a published tag is exactly the mutability this Action exists to argue against — on 19 March 2026 an attacker force-pushed malware over 76 of 77 release tags of aquasecurity/trivy-action — and the argument does not get an exemption for being inconvenient to its own author.

The floating v0 tag does move, which is the documented Marketplace convention; pin to v0.1.1 if you want the guarantee.

• This action is used across all versions by 44 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Plumber is an open-source CI/CD security scanner that uses a Rego policy engine to scan .gitlab-ci.yml and .github/workflows/*.{yml,yaml} files. It helps identify risky patterns and security gaps in CI/CD pipelines, providing comprehensive reports in various formats such as terminal, JSON, SARIF, GitLab SAST, PBOM, and CycloneDX.

Plumber automates security scanning for GitLab CI and GitHub Actions, offering a one-command solution to ensure secure and compliant pipelines. The tool is designed to work with the latest versions of GitLab and GitHub, providing real-time feedback on potential vulnerabilities in your workflows.

What’s Changed

0.4.16 (2026-07-24)

✨ Features

• output: restructure control output into clear Passed/Skipped/Failed sections (78b576f)

👷 CI/CD

• release: pin v0.4.15 refs [skip ci] (5ad6b54)

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Supply-chain Guard is an open-source tool designed to scan and analyze dependency chains across various package managers and ecosystems, including npm, Pypi, Cargo, Go, RubyGems, Composer, NuGet, Docker, Terraform, VS Code extensions, GitHub Actions, and repositories. It detects malware campaigns (e.g., GlassWorm, Vidar), fake AI tool repos, account takeovers, and numerous threat indicators across multiple lockfile formats. The action generates CycloneDX SBOMs, validates SLSA provenance, and correlates findings into attack-chain incidents for enhanced security awareness and remediation.

What’s Changed

[5.17.8] - 2026-07-24

Threat intel: jscrambler npm compromise, cPanel/WHM GitHub Actions abuse, and the Apex macOS infostealer

Added detection for three developer-targeted supply-chain campaigns whose concrete indicators were sourced from primary vendor write-ups (Socket, The Hacker News, OX Security, StepSecurity, safedep, Rescana) after the daily news scan surfaced them by name only.

• jscrambler npm compromise (2026-07-11). The jscrambler package (~15,800 weekly downloads) and four companion build plugins were hijacked and republished with a native Rust infostealer: a malicious preinstall hook in 8.14.0-8.17.0, then a self-executing dropper in dist/index.js and dist/bin/jscrambler.js from 8.18.0. The payload harvests AWS, GCP and Azure credentials, crypto wallets, browser data and AI-tool configs on Windows, macOS and Linux. Last clean release is 8.13.0, fixed in 8.22.0. The previous partial entry (8.14.0 only) is extended to the full malicious set.
• cPanel/WHM GitHub Actions abuse (2026-07-23). A legitimate developer’s ten Packagist packages had malicious dev-main versions injected with dozens of GitHub Actions workflow files that spin up runners, pull an architecture-specific Linux payload from 43[.]228[.]157[.]68, and scan for cPanel/WHM servers vulnerable to CVE-2026-41940 to harvest credentials, SSH material and cloud keys. Only the network and file indicators are ingested: the maintainer is a victim, so neither the account nor the bare package names are flagged.
• Apex macOS infostealer (2026-07-22). A postinstall dropper installs an AMOS-family macOS infostealer while installing a working forked coding agent as cover. npm removed @apexfdn/apex; the operator re-published the same payload as @copilot-mcp/apex about eleven hours later and churned twenty-plus versions in eight hours, so the packages are blocked by name rather than version.

Added

• Version-pinned jscrambler (8.14.0/8.16.0/8.17.0/8.18.0/8.20.0), jscrambler-webpack-plugin 8.6.2, gulp-jscrambler 8.6.2, grunt-jscrambler 8.5.2 and jscrambler-metro-plugin 9.0.2 in KNOWN_BAD_NPM_VERSIONS (src/ioc-blocklist.ts) and BUNDLED_FEED (src/threat-intel.ts), plus five SHA-256 payload hashes.
• 43[.]228[.]157[.]68 in KNOWN_C2_IPS, the DNS-callback subdomain f5b0b742-240a-4811-8a5b-b0ba6060685d[.]dnshook[.]site in KNOWN_C2_DOMAINS, and the Linux exploit payload SHA-256 in KNOWN_MALICIOUS_HASHES for the cPanel/WHM campaign, mirrored as BUNDLED_FEED entries.
• ^(@apexfdn\/apex|@copilot-mcp\/apex)$ to MALICIOUS_PACKAGE_PATTERNS (src/patterns.ts) plus bare-name BUNDLED_FEED entries for both packages.
• Campaign test coverage in src/__tests__/campaigns.test.ts for all three campaigns, including negative tests for the clean jscrambler 8.13.0 release.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Redflag is a cross-platform CLI that scans source files and Git history for secrets using regular-expression rules and heuristic Shannon entropy checks. It helps automate the detection of sensitive information, such as passwords and API keys, in both codebases and version control systems. The action integrates with GitHub Actions to perform secret scanning in CI pipelines.

What’s Changed

What changed

• Renamed the Action to Redflag Secret Scanner to satisfy the GitHub Marketplace unique-name requirement.
• Published the reusable Action at iammerus/redflag@v0.1.1.
• Updated the package version and README examples to 0.1.1.
• Kept scanner behaviour unchanged from v0.1.0.

Usage

- uses: actions/checkout@v4
  with:
    fetch-depth: 0
- uses: iammerus/redflag@v0.1.1
  with:
    git-history: "true"

Full changelog: https://github.com/iammerus/redflag/compare/v0.1.0...v0.1.1

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

MCP Trust Checker is a deterministic security scanner that assesses Model Context Protocol servers to ensure they are safe before connecting them to data. It uses the Capability-Flow Trust Model, an original algorithm, to evaluate servers based on their roles and behavior, identifying threats such as untrusted input ingress, sensitive data sources, and exfiltration paths, ensuring they do not pose a risk.

What’s Changed

Methodology mcptrustchecker-1.9 — unchanged. No weight, rule, gate or grade band moved, so every grade this release produces is identical to 1.9.0.

Added — a GitHub repository is now a scan target

mcptrustchecker scan upstash/context7
mcptrustchecker scan https://github.com/modelcontextprotocol/servers
mcptrustchecker scan owner/repo@v1.2.3

• Accepts owner/repo, owner/repo@ref, https/ssh github.com URLs, .git suffixes and /tree/<ref> links. Every other target shape is untouched — a scoped package, a path and a live URL still take their own branch, pinned by a regression test.
• Nothing is cloned, written to disk or executed. The archive is fetched over https from GitHub’s pinned hosts and unpacked in memory by the same bounded reader used for npm/PyPI artifacts: same size cap, same redirect re-validation, same zip-slip-safe entry paths, same unpacked-bytes ceiling. No git binary runs, so no hook can fire.
• A repository is not a released artifact and the scan says so: the surface is marked repo, verification is the repo tier (never source or vendor, which require publish provenance), and the archive’s own SHA-256 is recorded — “the default branch” is only reproducible against the bytes actually read.
• GITHUB_TOKEN or --github-token lifts GitHub’s anonymous rate limit; it is sent to api.github.com only and never survives a redirect.

Fixed — DNS rebinding could walk past the SSRF guard

isBlockedHost() compared the hostname as a string, so a public name whose A record points at loopback sailed through. That mattered most in OAuth, where discovery, registration and token endpoints are chosen by the target server’s own metadata.

• New isBlockedHostResolved() resolves the name and blocks it when any returned address is non-routable — a multi-record name is only as safe as its worst answer. Applied at the acquireHttp entry and on every fetch hop, so it covers redirects and every OAuth host.
• Private-range coverage widened: CGNAT (100.64/10), multicast (224/4) and reserved space (240/4). 100.63.x and 100.128.x stay routable.
• --allowed-hosts now overrides the private-host guard, which previously refused a host you had explicitly allowed — 100.64/10 is Tailscale’s range.
• The guard’s DNS lookup is time-boxed to 3s; dns.lookup takes no timeout.

Deliberate limits: an unresolvable name is not treated as blocked, and a sub-TTL rebind between lookup and connect still needs a custom dispatcher to close.

Install

npm i -g mcptrustchecker@1.10.0
mcptrustchecker --version     # 1.10.0 (methodology mcptrustchecker-1.9)

415 tests pass, including new ones pinning the repository parser and the resolved host guard.

Also live: the free online scanner — https://mcptrustchecker.com/scan

Full changelog: https://github.com/illiahaidar/mcptrustchecker/blob/main/CHANGELOG.md

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

droast is an opinionated Dockerfile linter that catches bad practices and reports them in a blunt manner. It supports various features such as parsing heredocs, handling parser directives, shell forms, BuildKit flags, Windows paths, and PowerShell, and provides real-time feedback with inline squiggles in VS Code and lint-on-save for Neovim users.

What’s Changed

• chore: release version 1.4.10 (6e9b82c)
• fix: handle dynamic images and modern Dockerfile syntax (cd2c37e)
• chore: release version 1.4.9 (a6b3c09)
• fix: resume interrupted release preparation (0759a95)
• fix: honor pip cache environment and synchronize release integrations (fa83565)
• fix: skip DF023 for final unnamed stage (266abe8)
• ci: skip VS Code publish without Marketplace token (bf12b21)
• chore: automate release preparation and tagging (3e686f6)
• chore: release version 1.4.8 (2b5b6bb)
• fix: detect unprotected RUN pipelines after set options (65b5c43)

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary: The GitHub Action ItzEmoji/aeroflare@v1 automates the process of building and pushing Nix packages to an OCI cache, providing a stateless, zero-infrastructure binary substituter. It supports pushing builds directly from CI without requiring additional tooling on the host machine. The action can build all outputs or specific ones based on changes or specified patterns, making it suitable for continuous integration workflows where Nix is used for packaging and caching dependencies.

What’s Changed

1.13.0 (2026-07-24)

Features

• changed builds sentinel (06ff04d)
• ci: add a changed builds sentinel that diffs derivations (2af89ef)

Bug Fixes

• ci: attribute the NUR commit to the app’s own bot (15ec767)
• ci: diff changed against the nearest evaluatable ancestor (25cd1c3)

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action installs a specific version of the stackit-cli tool on a runner, allowing developers to automate the setup and use of stackit-cli in their workflows. It supports fetching either the latest stable release or any specified semantic version, making it easy to integrate stackit-cli into CI/CD pipelines for deployment and management tasks.

What’s Changed

What’s Changed

🛠️ Dependencies

• chore(deps): update dependency undici to v8.9.0 by @renovate[bot] in https://github.com/jkroepke/setup-stackit-cli/pull/294

Full Changelog: https://github.com/jkroepke/setup-stackit-cli/compare/v1.2.91...v1.2.92

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of uploading SBOMs, SLSA attestations, and other artifacts to the Kusari Platform from a GitHub workflow. It simplifies integration by handling authentication credentials and provides options to generate SBOMs automatically or manually. Key features include capturing ingestion results for machine-readable IDs and ensuring components are mapped if needed.

What’s Changed

What’s Changed

• Update kusari-cli to v2.9.1 by @karlyanelson in https://github.com/kusaridev/kusari-ingest/pull/42

Full Changelog: https://github.com/kusaridev/kusari-ingest/compare/v4.8.0...v4.9.0

• This action is used across all versions by 107 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action is designed to automate the continuous localization process within a repository using the Lingo.dev platform. It helps teams manage translations efficiently and ensure consistency across multiple locales, reducing errors by leveraging AI-assisted tools and connecting directly to translation APIs.

What’s Changed

Patch Changes

• Updated dependencies [8c0393a]:
  • @lingo.dev/_sdk@0.17.3

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

lgtmaybe is a GitHub Action that reviews pull requests by analyzing code changes and security vulnerabilities, providing inline comments and summaries on GitHub. It uses an OpenAI-compatible model to generate feedback without checking out or running the code, focusing on logic errors, security risks, tests, documentation updates, performance issues, complexity, intent, and unnecessary complexity. The action can run in different presets for varying levels of thoroughness, including default fast mode and full audit mode for release branches.

What’s Changed

1.1.0 (2026-07-24)

Features

• diff: skip generated llms.txt/llms-full.txt files in review (#202) (b06086c)
• enable diagrams in starter workflows (#204) (157c47d)

Bug Fixes

• align action with v1 release (#209) (46e94b5)
• bound default review runtime (#208) (d21e4e8)

Documentation

• add Google Search Console verification file (#201) (d20b6d3)
• clarify Marketplace provider setup (#206) (089e74f)
• improve homepage feature showcase (#207) (ab9984f)
• show change diagram on homepage (#205) (64f58f7)

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the scanning of dependency lockfiles in a repository for known CVEs using vuln.mlab.sh. It automatically detects common lockfile types and checks each package against OSV and Sonatype OSS Index, providing detailed vulnerability reports in the job summary. The action can fail the build based on a specified severity threshold or report vulnerabilities without failing, with outputs for total vulnerabilities, vulnerable packages, and failure status.

What’s Changed

First public release. 🎉

Scan your lockfiles for known CVEs in CI/CD, powered by vuln.mlab.sh.

Features

• Auto-detects lockfiles: Cargo.lock, package-lock.json, npm-shrinkwrap.json, composer.lock, Gemfile.lock, go.sum, requirements.txt, mise.lock.
• Server-side parsing + scan against OSV & Sonatype OSS Index — no local install, no database.
• Severity threshold with fail-on (any · critical · high · medium · low · none).
• soft-fail mode to report without failing the build.
• Job summary table + inline annotations on vulnerable dependencies.
• Outputs: total, vulnerable-packages, failed.
• Optional API token (token:) for 25 scans/hour (vs 8/hour anonymous).

Usage

- uses: mlab-sh/vuln-scan-action@v1
  with:
    fail-on: high
    token: ${{ secrets.VULN_MLAB_TOKEN }}

See the README for all inputs.

• This action is used across all versions by 10 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automatically updates a specific Go module proxy cache when new tags are created, ensuring that the latest module versions are immediately available and documentation is updated on platforms like pkg.go.dev. It supports both standard and submodule version tags and allows customization of the proxy configuration, import path, and Go version used for building the module.

What’s Changed

1.1.32 (2026-07-24)

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The aer action automates running Apex tests locally, providing a fast feedback loop without using an org or deploy. It supports running unit tests with code coverage and executing anonymous Apex, as well as debugging Apex in interactive mode through VS Code or IntelliJ. The action can be installed via Homebrew, the Salesforce CLI, or manually downloaded and added to the PATH.

What’s Changed

Version v1.2.24

• Fix Generic And Array Class-literal Type Names

• Persist CartTestUtil.createCart Records And Widen ProcessException.AttachedToId

• Fix Infinite Recursion Resolving Methods When A Class Extends Its Own Inner Class

• Maintain WebCart Summary Totals As Platform Roll-Ups

• Support The Id.valueOf(String, Boolean) Restore-Casing Overload

• Accept Already-Qualified Class Names In Two-Argument Type.forName

• Correct Content SObject Describe Metadata

• Truncate Auto-Generated User CommunityNickname To The Field Length

• Replace Unknown-Typed Standard-Field Stubs When Feature Schemas Merge

• Support ConnectApi Output Representation Field Inheritance

• Gate Field-Name Auto-Feature Detection On An SObject Target

• Resolve Instance Calls To Inherited Instance Method Over Static Shadow

• Import Fields Declared Inline In An object-meta.xml File

• Fix Flow Formula Datetime And Merge-Field String Conversion

• Parse Custom Labels Through force-md To Accept Undeclared HTML Entities

• Report RelationshipOrder For Builtin Master-Detail Fields

• Guard Batch-update Audit-field Stamping On Field Existence

• Derive ContentVersion.Title From PathOnClient When Omitted On Insert

• This action is used across all versions by 16 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action, designed by oliversalzburg, automates the creation of semantic releases. It helps maintain consistency in version numbers and release processes across projects. The action supports both tagged builds and automatic releases on pushes or schedules, ensuring that versions are managed correctly without manual intervention.

What’s Changed

Chores (2)

• Rebuild entrypoint (Oliver Salzburg)
• Version bump 3.3.1 (Oliver Salzburg)

Features (1)

• Don’t set content-length header (Oliver Salzburg)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action generates directory listings for a repository’s files and directories and automates their automatic deployment to a specified branch, either within the same repository or an external repository using personal access tokens. The action is useful for creating static websites directly from a GitHub repository without additional setup, and it supports various configuration options such as excluding certain files, setting custom site URLs, and managing publishing directories.

What’s Changed

What’s Changed

• chore: improve styling by @pranabdas in https://github.com/pranabdas/directory-listing/pull/2

Full Changelog: https://github.com/pranabdas/directory-listing/compare/v1...v1.0.1

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action described in the README facilitates automated security assessments within cloud environments using Prowler, an Open-Source Cloud Security Platform. It simplifies the process of conducting real-time monitoring and customizable vulnerability scans, ensuring organizations can maintain a secure and compliant state across multiple cloud platforms.

What’s Changed

✨ New features to highlight in this version

Enjoy them all now for free at https://cloud.prowler.com

🎫 Grouped Jira workflows

[!NOTE] This feature is available exclusively in Prowler Cloud and Prowler Private Cloud with a subscription.

Selected Findings, Finding Groups, and mixed selections can now be sent to Jira. When you select multiple findings, choose between one grouped issue or separate issues. Generated issues keep their Prowler context with deep links and filter details, while the UI provides clear dispatch and failure feedback.

Read more in our Jira integration documentation.

🕸️ Cleaner Attack Paths results

[!NOTE] This feature is available exclusively in Prowler Cloud and Prowler Private Cloud with a subscription.

Prowler Cloud now records which built-in Attack Paths queries returned data at the end of each scan. The query selector hides confirmed-empty queries for the selected scan, so you can focus on paths that exist without opening blank graph views. Errored, unknown, and parameterized queries remain available when they still require investigation or input.

All Attack Paths queries are now published on Prowler Hub, where you can browse the full catalog.

Read more in our Attack Paths documentation.

🧑‍🏫 New Tutorials: Connect Your AI Agents to Prowler Cloud

[!NOTE] For this feature you need a Prowler Cloud API key so this is available exclusively in Prowler Cloud and Prowler Private Cloud with a subscription.

New tutorials walk you through connecting your own AI agents to Prowler Cloud, so they can query your security posture and act on it programmatically.

Read more in our AI agents documentation.

☁️ Simpler OCI provider setup

OCI provider credentials no longer require a region. Existing clients can still send the legacy region field for compatibility, but the API ignores it before storing credentials or starting a scan. This removes an unnecessary step from OCI onboarding.

Read more in our OCI documentation.

🔍 Checks

AWS

• sagemaker_notebook_instance_no_secrets scans the OnCreate and OnStart lifecycle scripts of SageMaker notebook instances for hardcoded API keys, passwords, tokens, connection strings, and other secrets. Thanks to @kiranrajsg!

Read more in our AWS documentation.

Explore all AWS checks at Prowler Hub.

🔐 Security

• Integration responses and operations now respect provider visibility, preventing hidden-provider disclosure and blocking unauthorized attachment, connection checks, Jira dispatches, edits, and deletion.
• Next.js was updated from 16.2.9 to 16.2.11, patching four high-severity and five medium-severity vulnerabilities.
• The unused npm CLI was removed from the UI container image, eliminating the bundled node-tar CVE-2026-59873 and reducing exposure to future bundled npm vulnerabilities.
• Vitest and its browser packages were updated from 4.1.8 to 4.1.10, resolving the critical @vitest/browser file-access permission bypass. These are development dependencies and have no runtime impact.
• Kubernetes kubeconfig validation now blocks legacy auth-provider.config.cmd-path command authentication, closing a command-execution bypass.
• next-auth was updated from 5.0.0-beta.30 to 5.0.0-beta.32, patching two critical Auth.js advisories: existence-based authorization checks that could fail open when a provider is misconfigured, and a homoglyph @ bypass in email address normalization. The bump also pulls in the patched @auth/core 0.41.3 transitively.

🙌 External Contributors

Thank you to our community contributors for this release!

• @kiranrajsg: AWS sagemaker_notebook_instance_no_secrets check (#11843)
• @owenchenxy: Alibaba Cloud SSH and RDP security group checks now handle capitalized Policy="Accept" values correctly (#12049)
• @rsaladra: S3 bucket name validation no longer raises an invalid escape sequence SyntaxWarning at startup (#12041)
• @SujayKulkarni-2211 - Updated the AWS check count in the README (#12011)

UI

🚀 Added

• Finding Groups and grouped selections can be sent to Jira in Cloud with deep links, filter chip display, and Jira feedback toasts (#12001)
• In Prowler Cloud, the Attack Paths query selector now lists only queries that returned data for the selected scan, hiding empty ones (#12010)
• Overview banner linking to the AI agents documentation, shown next to the Lighthouse AI banner in Cloud and full width on self-hosted deployments (#12074)

🐞 Fixed

• Findings Severity Over Time chart Y-axis labels no longer overflow for large findings counts (#11545)
• UI Sentry alerts now suppress non-actionable warnings and expected API/control-flow noise while preserving actionable runtime failures (#11665)
• OCI provider E2E tests no longer require or submit a region when adding or updating credentials (#11741)
• Billing navigation is hidden when Cloud billing is disabled, including Enterprise deployments (#12047)
• AWS Organizations setup modal now shows the “Enter a valid Organizational Unit or Root ID” hint in the error color, clarifying why the deployment button is disabled (#12063)
• Sidebar logo top spacing in the main app sidebar (#12066)
• Contextual Cloud upgrade modal content remains stable throughout the closing animation (#12067)
• Tenant switches now refresh session user permissions for the selected tenant (#12087)

🔐 Security

• Removed the unused npm CLI from the UI container image, eliminating the bundled node-tar CVE-2026-59873 (and future bundled-npm CVEs); the image builds with pnpm via corepack and does not use npm (#12065)
• Bumped vitest and @vitest/browser, @vitest/browser-playwright, @vitest/coverage-v8 from 4.1.8 to 4.1.10, resolving the critical @vitest/browser Browser Mode file-access permission bypass (GHSA-p63j-vcc4-9vmv) flagged by pnpm audit; dev dependencies only, no runtime impact (#12077)
• Kubernetes credential forms now reject kubeconfig files using legacy auth-provider.config.cmd-path command authentication (#12091)
• Next.js from 16.2.9 to 16.2.11, patching 4 high- and 5 medium-severity vulnerabilities (#12093)
• next-auth from 5.0.0-beta.30 to 5.0.0-beta.32, patching 2 critical Auth.js advisories (GHSA-8fpg-xm3f-6cx3 fail-open auth checks, GHSA-7rqj-j65f-68wh email homoglyph bypass) (#12108)

API

🔄 Changed

• OCI provider secrets no longer require region; legacy region input is accepted for backwards compatibility but ignored before storing or scanning (#11741)
• Compliance overview ingest now runs in a single transaction per scan with a configurable COPY batch size (DJANGO_COMPLIANCE_COPY_BATCH_SIZE, default 2000), reducing write pressure on the database (#11875)

🐞 Fixed

• Scan findings now recover resources missing from the in-memory cache after resource pre-resolution, preventing valid findings from being skipped (#12002)
• Tenant-wide integrations that are not attached to any provider, such as Jira, are now visible and manageable by roles with manage_integrations and without unlimited visibility (#12060)
• Output generation now removes the scan’s temporary output directory before writing, so a re-run of the task for the same scan (e.g. broker redelivery after a worker is killed mid-run) no longer appends to the previous run’s files and duplicates finding rows in the exported CSV and other outputs (#12097)

🔐 Security

• Integration responses no longer disclose providers outside the visibility of the role, including the resources sideloaded through ?include=providers (#12060)
• Integration connection checks, Jira issue type lookups and Jira dispatches now resolve the integration through the provider visibility of the role instead of the whole tenant (#12060)
• Roles without unlimited visibility can no longer attach an integration to providers they cannot see, nor edit or delete an integration bound to them (#12060)
• Kubernetes kubeconfig validation now rejects legacy auth-provider.config.cmd-path command authentication in Prowler Cloud/API (#12091)

SDK

🚀 Added

• sagemaker_notebook_instance_no_secrets check for AWS provider, scanning SageMaker notebook instance lifecycle configuration scripts (OnCreate and OnStart) for hardcoded secrets such as API keys, passwords, tokens, and connection strings (#11843)

🔄 Changed

• Jira output rendering supports grouped Finding Group issues with caller-provided links and capped or uncapped finding copy (#12035)

🐞 Fixed

• Fix invalid escape sequence SyntaxWarning raised on startup by the S3 bucket name validation regex (#12041)
• Alibaba Cloud SSH and RDP security group checks no longer produce false negatives when allowed rules use capitalized Policy="Accept" values (#12049)

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action allows developers to execute a command in the specified ECS Fargate Service. It automates database migrations by running scripts directly within the service containers, reducing the need for manual intervention and improving efficiency. The action handles multiple container services by requiring a specific container name when necessary, ensuring consistent execution across all tasks within the service.

What’s Changed

container_name input (optional) → passed to ExecuteCommandCommand.container. Backward-compatible; omitting it keeps today’s behavior.

desiredStatus: RUNNING filter on ListTasks + explicit error if no running task (previously could grab a stopped/undefined task).

Exit-code detection — wraps the command so it echoes ECS_EXEC_EXIT=$? and scans the streamed output; the step now fails on a non-zero exit or a dropped session. This closes the silent-failure hole the raw CLI approach also had.

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action automates the deployment, testing, and release management of Salesforce projects using the @sfdt/cli tool. It simplifies the process by providing features such as interactive workflows with preflight validation, automated release manifests, parallel Apex test execution, AI-powered fix plans for code and tests, and CI/CD pipeline templates for GitHub, GitLab, Azure, and Bitbucket. The action also supports multi-package projects and provides a local web dashboard for monitoring and comparing org states.

What’s Changed

What’s Changed

• chore: promote develop → main (extension v0.8.1) by @scoobydrew83 in https://github.com/scoobydrew83/sfdt/pull/268
• chore: promote develop → main — CLI v0.19.0 + extension v0.9.0 by @scoobydrew83 in https://github.com/scoobydrew83/sfdt/pull/269

Full Changelog: https://github.com/scoobydrew83/sfdt/compare/v0.18.2...v0.19.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Reelier automates the verification of AI agent tool-call workflows against new dependencies, ensuring consistency and reliability in software development processes. By replaying recorded runs at zero LLM cost and diffing them, Reelier helps identify any drift between the original run and the updated environment, providing clear receipts for audits and approvals before merging changes.

What’s Changed

Adds a Dependabot/Renovate bump-safety check: replay your recorded agent tool-call skills live against a bumped dependency at 0 LLM tokens, and fail the PR check on the exact step that drifted.

• New wrap input to front your own MCP server / tool process (the case that exercises a bumped SDK your tool code imports).
• Copy-paste recipe: .github/workflows/reelier-bump-check.yml.
• Honest scope: at max-level 0 the replay never calls an LLM, so it verifies dependency / MCP-tool-call behavior, not model upgrades.

Also works unscoped from bump PRs, as a scheduled drift check or a plain “replay this skill on every PR” gate. See the README.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Argus is an AI code review tool that optimizes recall to find more real bugs while keeping false positives manageable. It runs multiple specialized AI reviewers in parallel and uses an evidence-based curator to verify findings before posting review comments on GitHub pull requests. The planner briefs every reviewer up front, lenses focus on specific problem domains, and the curator merges duplicates and only dismisses issues with cited quotes from the diff.

What’s Changed

What’s Changed

• Isolate lens failures and size context per model by @sibinms in https://github.com/sibinms/argus/pull/46
• Bump version to 1.2.28 by @sibinms in https://github.com/sibinms/argus/pull/47
• Release v1.2.28 by @sibinms in https://github.com/sibinms/argus/pull/48

Full Changelog: https://github.com/sibinms/argus/compare/v1.2.27...v1.2.28

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary: SSG is a fast Go-based static site generator that converts Markdown with YAML frontmatter into a complete website. It supports various features such as built-in themes, templates engines, SEO metadata, image processing, and deployment to multiple platforms like GitHub Pages, Netlify, Vercel, and more. The action automates content generation, rendering, and deployment processes, making it ideal for blog creation, documentation, and other static site needs.

What’s Changed

Installation

Quick Install (Linux/macOS)

curl -sSL https://raw.githubusercontent.com/spagu/ssg/main/install.sh | bash

Package Managers

• Homebrew: brew install spagu/tap/ssg
• Snap: snap install ssg
• Debian/Ubuntu: Download .deb file below
• Fedora/RHEL: Download .rpm file below

Checksums

See checksums.sha256 for file verification.

📖 Full documentation: https://github.com/spagu/ssg#readme

What’s Changed

• 1.8.13 — workers: plural workers, cookie-consent, comments, republish-trigger, config includes by @spagu in https://github.com/spagu/ssg/pull/50

Full Changelog: https://github.com/spagu/ssg/compare/v1.8.12...v1.8.13

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action packages a Kingdom Come: Deliverance loose-file mod into game-ready .pak files and a release .zip. It solves the problem of loading mods by ensuring that the paks do not have unnecessary modification-time metadata, which 7-Zip / WinRAR / WinZip add. The action supports both compression methods (deflate or store) and provides outputs for the mod ID and version, allowing users to manage their releases from the mod.manifest file.

What’s Changed

Package a Kingdom Come: Deliverance (KCD1/KCD2) loose-file mod into game-ready .pak files and a release .zip. Runs as a GitHub Action or as a local Python tool.

Highlights

• CryPak-safe paks: writes ZIPs with zero per-file modification-time extra fields (NTFS 0x000A / extended-timestamp 0x5455), the metadata that makes a plain renamed 7-Zip/WinRAR archive fail to load in KCD2.
• Correct pak layout: Data/** except Data/Levels/** -> Data/<modid>.pak; Localization/<Lang>/** -> Localization/<Lang>.pak; Data/Levels/<level>/** -> Data/Levels/<level>/<modid>.pak; mod.manifest and mod.cfg are kept loose.
• Compression: deflate (default) or store.
• Outputs: archive, deploy-dir, modid, version (read from mod.manifest).

Usage

- uses: actions/checkout@v7
- uses: tkhquang/kcd-pak-action@v1
  with:
    mod-dir: src        # folder containing mod.manifest
    compress: deflate   # or 'store'

See the README for both the tag-triggered and manifest-driven release workflows.

Requirements

• Python 3.8+ on the runner (present on GitHub-hosted runners). Also runs locally: python scripts/pak_kcd_mod.py <mod-src-dir>.

License

0BSD.

• This action is used across all versions by 10 repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

NeuroLink is the universal AI integration platform that simplifies integrating 30+ AI providers and models into applications. It provides a TypeScript-first interface that supports various functionalities such as text streaming from LLM providers, handling different output modes (avatar and music), and deploying with enterprise features like Redis memory and failover. The action automates these tasks and streamlines the integration process for developers.

What’s Changed

10.5.2 (2026-07-24)

Bug Fixes

• (core): make supportsTools model-aware via the model registry (c4137f3)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action checks AI agent configurations to catch risky changes such as safety instructions removal, dangerous tool permissions addition, and model tier swaps. It automates these checks before merging changes into a repository, providing detailed reports and alerts for sensitive configuration modifications. The action supports free and Pro tiers with varying features and reporting options, making it suitable for both casual users and organizations requiring more advanced observability.

What’s Changed

Initial release: git-native check command, GitHub Action, and Pro tier (Gumroad-verified license: full per-file detail, model cost-tier detection, SARIF output).

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

pr-sage is an AI-powered pull request reviewer designed to minimize noise and follow team conventions. It reviews only the most recent changes since the last review, uses fingerprinting to eliminate duplicate comments, and can enforce severity levels and quality gates in PRs. The action supports multiple providers (Anthropic, OpenAI, Gemini) and can be used locally for pre-push reviews without needing a GitHub token.

What’s Changed

Review safety, onboarding, and operations improvements.

Highlights

• Finding lifecycle — follow-up reviews report which earlier findings were fixed and which remain unresolved
• Coverage honesty — every summary reports what was actually reviewed; partial reviews never auto-approve, and --fail-on-incomplete turns incomplete coverage into a CI failure
• Verification controls — --verify-provider / --verify-model to cross-check findings with a different model, --verify-failure abort|keep|drop
• pr-sage doctor — diagnose config, keys, and workflow wiring in one command
• GitHub Check Run annotations (--check-run)
• Hardened generated workflow — config loaded from the trusted base commit, concurrency cancellation, self-hosted runner + openai-base-url support for fully local reviews
• Scheduled benchmark workflow

Since v0.5.0 (v0.6.0)

pr-sage init 30-second setup wizard, draft/WIP/skip-label skip rules, maxTokensPerRun cost guard, monorepo paths scoping, locale: auto detection, README demo GIF.

npx pr-sage init   # 30-second setup

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 90 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The @linear/release-action GitHub Action automates the synchronization of deployments with Linear releases by integrating CI/CD pipelines into Linear’s release management system. It scans commits for Linear issue identifiers and pull request references, creating or updating releases in Linear automatically. This action helps teams manage releases more efficiently by ensuring that all related issues are linked to the correct releases.

What’s Changed

What’s Changed

• Fix release PR body wording for drifted CLI version by @RomainCscn in https://github.com/linear/linear-release-action/pull/53
• Release v0.14.6 by @RomainCscn in https://github.com/linear/linear-release-action/pull/54

Full Changelog: https://github.com/linear/linear-release-action/compare/v0.14.5...v0.14.6

• This action is used across all versions by 0 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Git Velocity is a GitHub Action that analyzes your code contributions and generates a game-like dashboard to track developer velocity. It automates the process of analyzing repositories, calculating scores and achievements, and providing insights into team activity patterns. The action supports local Git analysis, caching, and authentication options, making it fast and flexible for developers to monitor their contribution velocity effectively.

What’s Changed

Changelog

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action mirrors repository milestones into Jira epics, ensuring one-way synchronization. It maintains an idempotent and safe operation, handling both scheduled runs and manual triggers. The action updates the epic’s summary, description, due date, and labels as needed, while preserving existing manual edits in Jira.

What’s Changed

GitHub Milestones → Jira Epics — v1.0.0

Mirror a repository’s GitHub milestones into Jira Cloud epics — one-way, idempotent, safe to run on a schedule.

Core sync ♻️

• One epic per milestone, matched by a structural gh-ms- label via exact JQL lookup — survives title edits and manual changes.
• Idempotent upsert: creates epics for new milestones, updates existing ones, and performs zero writes when nothing changed.
• Change detection: summary, description, and duedate are written only when they actually differ — no noisy Jira history or watcher spam.

Field mapping 🔗

• Summary = milestone title verbatim.
• Rich ADF description: “do not edit” notice, a closed/total issues closed header, the milestone description, and Open/Closed issue lists — with pull requests excluded from both counts and lists, sorted deterministically.
• Due date set only when the milestone has one, and never cleared afterward.
• Start date never touched — a manually set start date always survives.
• Labels reconciled as a superset: adds gh-ms- + your configured labels, never prunes labels a human added in Jira.
• Status sync by status category (done/new) — closes/reopens epics with milestone state, respects a manually set In Progress, and works across renamed or non-English statuses.

Configuration & safety 🔒

• Fully repo-agnostic: everything (project key, epic type id, extra labels, tokens) is an action input — nothing hardcoded.
• Dry-run mode to preview every create/update/transition without writing.
• only input/flag to scope a run to a single milestone (near-real-time on milestone events).
• Fail-fast auth preflight with an actionable message (including the classic-vs-scoped API-token pitfall), and Jira error responses surfaced in full — no opaque 400s.
• Bounded retry/backoff on rate limits and transient 5xx.
• Clear run summary and exit codes; outputs created / updated / unchanged / skipped.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The lgtmaybe GitHub Action reviews PRs by analyzing code changes and surrounding lines to identify potential issues such as logic errors, security vulnerabilities, and missing tests. It provides inline review comments and a summary of findings, helping reviewers focus on the most critical changes in context. The action supports multiple review levels (info through critical) and can be configured to run different presets based on branch status.

What’s Changed

1.0.0 (2026-07-24)

Features

• action: post reviews as a GitHub App via app_id/app_private_key inputs (#197) (88ec1e4)

Documentation

• optimise the docs site for SEO and LLM crawlers (#200) (d46078f)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The move-test-gen GitHub Action is an Agent Skill that automatically generates edge-case test suites for Sui Move functions. It focuses on covering various edge cases such as boundary values, arithmetic edges, access control issues, state machine problems, and economic concerns, providing a comprehensive set of tests to ensure robustness in Move smart contracts. The action can be used to generate tests based on user requests or automatically check the coverage of existing tests to identify potential gaps.

What’s Changed

What’s new in v1.3.0

MOV-002 and MOV-004 now use a lightweight Move parser instead of line-level regex. The parser tracks variable types through declarations, casts, and destructuring — so it knows that numerator1 declared as liquidity_u256 << RESOLUTION is u256, not u64. This killed most of the false positives that required naming-convention heuristics in v1.2.0.

MOV-004: unsafe downcast (new rule)

Flags (expr as u64) from u128/u256 without a preceding overflow assert. Move’s as silently truncates — unlike arithmetic ops which abort. Found 3 true positives in Bucket Protocol (pending reward/debt calculations downcast mul_factor_u128 results without checking), which led to the fix PR below.

Parser integration (MOV-002)

Turbos CLMM went from 14 findings to 3 after the parser was wired in. Variables cast to u256 on earlier lines are now tracked through the function body instead of relying on _u256 suffix matching.

Bucket Protocol fix PR

MOV-002 flagged collateral_raw_value * pow(10, decimal_diff) in Bucket’s value conversion functions — mul_factor promotes to u128 internally, the scaling multiplication right after undoes it. Submitted a fix: https://github.com/Bucket-Protocol/v1-core/pull/12

Validation

Infrastructure

• scripts/move-parser.mjs — function-level parser: signatures, params with types, variable declarations, cast tracking, assert locations
• gate-selftest: 11/11 cases, MOV-004 pins added (unsafe flagged, assert skipped, library skipped, small field skipped, test skipped)
• Rules: 4 (MOV-001 through MOV-004)

Diff from v1.2.0: https://github.com/mehvetero/move-test-gen/compare/v1.2.0...v1.3.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Totem is a tool that keeps project lessons and rules in the repository itself, ensuring architectural integrity by preventing re-inventing standard patterns. It uses a file-based toolkit with plain markdown lessons, a queryable knowledge index derived from them, and compiled lint rules enforced by a local, deterministic zero-LLM linter. This approach reduces friction and improves code quality by preventing common architectural mistakes.

What’s Changed

Cohort-link bump (no direct package changes). See .changeset/config.json for the fixed-cohort definition.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

MotionScore Guard automates animation performance auditing for pull requests by loading pages in a real browser on GitHub runners and grading them with S to F tiers. It supports free access for any repository but provides a paid plan with token-based gating that fails builds if page grades fall below specified thresholds. The action can audit specific paths and upload reports, making it easy to integrate into workflows.

What’s Changed

Runs motionscore CLI 0.1.2.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The repro-check GitHub Action automates the process of running and troubleshooting old research code, helping to identify and fix issues such as missing files, outdated dependencies, and removed APIs. It provides a runnability scaffold for reproducing computational papers by finding scripts, attempting to run them, and applying known fixes until they either run successfully or provide detailed information on where they stopped and what needs to be done next.

What’s Changed

--ai-suggest asks an LLM to draft a diagnosis + suggested fix on a hand-off, using YOUR OWN ANTHROPIC_API_KEY or OPENAI_API_KEY (your machine, your bill). Ships with no key of its own; no key set = feature off, no fallback. The suggestion is flagged, never auto-applied, and never counted as a fix. pip install --upgrade repro-check — https://pypi.org/project/repro-check/0.11.0/

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action, linkml-scala-action, automates the validation and generation of LinkML schemas in CI using Node.js. It leverages the @neverblink/linkml npm package to validate schemas by linting them or generating various output formats such as JSON Schema, SHACL, RDF/SDLC, and Scala classes. Key features include inline annotations for GitHub checks, fast execution times, and support for importing additional schema files. The action is designed to be lightweight and compatible with any Node.js runner environment.

What’s Changed

Tracks linkml-scala v0.11.1.

Bundles @neverblink/linkml@0.11.1.

• Pin uses: NeverBlink-OSS/linkml-scala-action@v0.11.1 for reproducibility.
• Pin @v1 for automatic patch/minor updates.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

AI Harness Doctor is a tool that audits AI harness files to ensure they are well-organized, up-to-date, and effective. It helps teams consolidate scattered guidance into a single AGENTS.md file while measuring improvements in agent answers, reducing latency, and optimizing costs. The action checks for inconsistencies, overlapping instructions, declaration-vs-code mismatches, and conflicts across various agent-config files.

What’s Changed

What’s Changed

• fix(drift): stop flagging GitHub Actions uses references as missing paths by @NieZhuZhu in https://github.com/NieZhuZhu/ai-harness-doctor/pull/332

Full Changelog: https://github.com/NieZhuZhu/ai-harness-doctor/compare/v1...v1.16.3

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of connecting to a NetBird network as an ephemeral peer, enabling access to internal services or databases over a secure WireGuard mesh. It allows users to configure various parameters such as setup keys, management URLs, and DNS settings to facilitate seamless connectivity. The action also provides an option to wait for the local NetBird DNS resolver to be ready before proceeding with subsequent steps, which is useful for ensuring proper resolution of private resources.

What’s Changed

What’s Changed

• Add dns-hosts so the action waits until listed hostnames resolve to private IPs only (NetBird CGNAT / RFC1918), avoiding public DNS wins that call APIs from outside the mesh (e.g. 403).
• Clarify split-horizon DNS wait behavior in README and action inputs.
• Simplify private-IP checks in wait-for-dns.sh.

Full Changelog: https://github.com/NomisCZ/netbird-cli-action/compare/v1.0.0...v1.1.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The action generates directory listings for GitHub Pages and automates the deployment process using GitHub Actions. It supports generating listings from a specified folder, excluding files or directories, and deploying them to either the same repository’s gh-pages branch or an external repository. The action can also customize various aspects such as site URL, base path, site name, footer text, and commit messages.

What’s Changed

What’s Changed

• major refactoring by @pranabdas in https://github.com/pranabdas/directory-listing/pull/1

New Contributors

• @pranabdas made their first contribution in https://github.com/pranabdas/directory-listing/pull/1

Full Changelog: https://github.com/pranabdas/directory-listing/commits/v1.0.0

• This action is used across all versions by 3 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the deployment of Cloudflare Workers through GitHub Actions, providing a consistent workflow for preview, dry-run, and production deployments. It leverages Wrangler to manage worker deployments and outputs relevant information about the deployment to the GitHub Actions job summary. The action handles authentication credentials and ensures that only qualified users can deploy production workers.

What’s Changed

1.1.0 (2026-07-21)

Features

• support package-local Wrangler (#5) (a282705)

• This action is used across all versions by 7 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary:

rumdl is a high-performance Markdown linter and formatter written in Rust. It offers over 77 lint rules, automatic formatting with the --fix option, and support for multiple Markdown flavors. The action provides detailed error reporting and is optimized for speed, making it suitable for use in CI/CD pipelines and editor integrations.

What’s Changed

Added

• mojibake: new rule MD083 for mojibake detection (#753) (552842b)

Fixed

• lint_context: prevent panic when HTML tag window splits a UTF-8 char (d14b252)

Downloads

Installation

Using uv (Recommended)

uv tool install rumdl

Using pip

pip install rumdl

Using pipx

pipx install rumdl

Direct Download

Download the appropriate binary for your platform from the table above, extract it, and add it to your PATH.

• This action is used across all versions by 5 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Bernstein is a deterministic orchestrator for CLI coding agents that schedules tasks in plain Python and ensures reproducibility and auditability. It uses an always-on lineage spine to track changes and an HMAC-chained audit log to verify results offline, making it suitable for environments with strict requirements on transparency and repeatability.

What’s Changed

A bug-fix release. This wave closes the defects surfaced by running 3.8.2/3.8.3 end to end across cluster mode, the task lifecycle, and the verifiability core: multi-machine setups that could never complete a task, runs that declared success without a deliverable, and audit/lineage surfaces that verified less than they claimed. Nothing here changes a public API; upgrade in place.

Cluster mode

• Worker registration works with a single credential: the raw cluster secret (or the operator API token) now clears both auth layers, which previously accepted disjoint credentials with no issuance surface for the inner one (#2805). The cluster secret is deliberately barred from operator-only admin endpoints.
• A tokenless or misconfigured worker join now fails fast with an actionable message naming --token, BERNSTEIN_AUTH_TOKEN, and the auto-generated token path, instead of silently retrying a 401 every 5 seconds forever; the deployment docs document the real join flow (#2802).
• Remote task completion is possible: the completion URL respects BERNSTEIN_SERVER_URL instead of hardcoding localhost, and the worker loop reports terminal state for every reaped process (#2808).
• Dead workers are detected and their claimed tasks requeued: the node reaper now always runs, and each claim records the owning node (#2801).
• Workers honor the task’s model/adapter selection instead of discarding it and refusing non-Claude adapters; a --model flag plus adapter profiles resolve sensible defaults (#2804).

Task lifecycle and run correctness

• A suspiciously fast clean agent exit is no longer auto-completed into “done” with an empty diff; it routes to retry and surfaces as clean_exit_unverified (#2810).
• Retry storms are capped: backoff and fail-counts follow the retried task lineage, so a structurally dead spawn terminates within the cap instead of ballooning a 5-step plan into ~97 task records (#2806).
• Resume after stop no longer declares a run complete while dropping open tasks: stopped sessions persist their open work and re-plan on resume (#2798).
• A supervisor watching an agent stuck at spawn-time “starting” now escalates through SIGTERM/SIGKILL and requeues the task instead of rewriting the shutdown signal forever (#2796).
• The CLI no longer opens a browser dashboard that will greet the operator with a 401; it probes first and prints the working alternatives (#2794).
• A manager/planning spawn that would run unsandboxed in the operator checkout is refused up front, and writes that escape the worktree are contained at reap time instead of landing untracked in the operator’s checkout (#2793).
• An explicit --sandbox override that cannot be honored fails loudly instead of degrading to host execution (#2847).
• --dry-run validates the seed through the same parser as a real run and warns on unparseable backlog files (#2849).

Audit chain, lineage, and replay

• The task server and spawner now resolve the same audit HMAC key, and appends hold a cross-process lock, so a mixed-writer chain verifies instead of failing on key mismatch; concurrent writers serialize into one verifiable chain (#2791).
• Ambient host BAGGAGE/TRACEPARENT env (foreign Sentry/OTEL ids) is no longer sealed into lineage hashes; only orchestrator-owned context is recorded (#2787).
• A lineage chain that contains only the journal seal no longer passes verification silently: bernstein lineage verify reports SEAL ONLY and exits non-zero when no artifact provenance exists (#2789).
• Deterministic replay refuses to run live: activating replay for a run with no recording aborts before any agent spawns instead of silently executing against the world (#2790).
• Nightly adapter-canary receipts are anchored into the HMAC audit chain as documented (#2850), chronic skip verdicts escalate to issues, and a stale or absent last-green is surfaced (#2845).
• Qwen CLI-adapter runs record real token usage in the cost ledger (#2851).

Cloud and Cloudflare adapter surfaces

• The cloudflare and codex_cloudflare adapters refuse fast with actionable errors instead of pretending: one had no worker-trigger path (every task timed out with zero artifacts), the other targeted a REST API that does not exist; both are marked experimental in docs (#2782, #2783).
• bernstein cloud init scaffolds a runnable free-tier worker, cloud deploy no longer points at a template the wheel does not ship, dead-endpoint errors are actionable instead of raw tracebacks, and the template declares no paid bindings (#2784).

CLI and dashboard

• Operator-surface defects from 3.8.2 testing: no more .claude/mcp.json churn on every run, --fresh works on run, stop --force reports only confirmed-terminated processes, ps sees a live leaf process, dry-run previews are populated, and the dashboard stops mislabeling models (#2800).
• worker --adapter accepts every registered adapter (same registry as run --cli), --idle documents that it forces the mock backend internally, start appears in --help, and container worker health is documented (#2807). Remaining wiring-level items are tracked in #2874.

CI and maintenance

• The post-CI dispatcher boots again: the auto-release call granted less than the callee requested and every dispatcher run since 2026-07-19 ended as a silent startup_failure — auto-release, auto-heal, CI-fix, and bisect were all dead (#2866). A regression test now asserts caller grants cover every callee job request (#2867).
• Hosted SonarQube, GlitchTip, Dependency-Track, and the operator Telegram alert integrations are retired to simplify operations (#2859, #2863); the observability snapshot runs on manual dispatch only (#2856).

Thanks

Thanks to the external contributors in this release:

• @AnayGarodia — centralized the artifact constants and typed the wire payloads so the task and adapter surfaces share one definition (#2838)

Full Changelog: https://github.com/sipyourdrink-ltd/bernstein/compare/v3.8.3...v3.8.4

• This action is used across all versions by 36 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action setup-unirtm automates the installation and configuration of UniRTM, a runtime and tools manager. It intelligently detects the best install method based on available runtimes and supports multiple methods such as npm, pip, GitHub Releases, and go installs. The action also offers features like caching using Handlebars templates and supports GitHub proxy configurations for restricted networks.

What’s Changed

0.6.0 (2026-07-24)

Features

• add support for concurrent blob downloads in artifact caching (9b9d1bb)

Bug Fixes

• correct pip package name and remove legacy Harden Runner (190f867)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action uses dlp-patterns to scan a codebase for secrets, PII, and sensitive data. It helps detect over 50 categories of sensitive information and provides detailed reports on findings, including the number and severity levels. The action can be configured to focus only on API keys and credentials or to scan everything, including PII. Users can set thresholds for severities to trigger different actions in CI/CD pipelines.

What’s Changed

Scan your codebase for secrets and PII in CI — zero configuration needed.

• Detects 50+ categories: AWS/GitHub/Stripe/Slack keys, JWTs, private keys, DB connection strings, SSNs, credit cards, and more
• Entropy gating and Luhn validation to minimise false positives
• secrets-only mode for fast source code scanning
• Configurable fail-on severity: critical, high, medium, low
• Outputs findings-count and highest-severity for downstream steps

Powered by dlp-patterns — Apache 2.0

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This action is a simple testing framework for Bash scripts that automates the process of writing and running tests. It focuses on providing developers with a lightweight, fast testing experience with over 70 assertions across various families, including equality, strings, exit codes, numeric checks, arrays, file/directory permissions, JSON assertions, date comparisons, duration checks, snapshots, and test doubles. The framework is designed to be user-friendly and developer-centric, offering interactive learning through the learn command and extensive documentation at bashunit.com.

What’s Changed

✨ Improvements

• Branch coverage now reports loop constructs (while/until/for/select): the loop body is a single-arm branch, marked covered only when the loop ran at least once, so a never-entered zero-iteration loop surfaces as an uncovered branch (#855)

🛠️ Changes

• Core string assertions (assert_contains/assert_not_contains, assert_matches/assert_not_matches, assert_string_starts_with/assert_string_ends_with and their negations) no longer fork a subshell per call to join their arguments; a fork-free join with identical behaviour replaces it (#844)
• The array, date, duration, json, files and folders assertions now resolve their failure label through the fork-free slot helper instead of a per-call command substitution — same labels, fewer forks
• Parallel test workers name their per-test result file by a per-suite ordinal instead of mktemp + mv, removing two forks per test (plus the echo \| tr \| sed arg sanitizing for data-provider tests) with identical result aggregation (#851)

👥 Contributors

• @Chemaclass

Checksum

SHA256: 151f3647964d53d3f5a7065c141790fc1b66ea3039024c80ed09b3a9602064a2

Full Changelog: 0.42.0…0.43.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The MCP Test Harness is a CI tool that automates and tests the correctness of MCP servers. It helps teams ensure their AI server features do not break silently and provides evidence of compliance with standards like JUnit, SARIF, and HTML reports. By using this action, developers can quickly verify the reliability and performance of their MCP servers before they are deployed, reducing costs associated with quality assurance and improving overall trust in their codebase.

What’s Changed

mcp-test-harness 3.0.9

CRA evidence packaging release: SBOM on publish, VDP security policy, optional --cra-output CRA reporter, and version alignment across core + 17 vendor shims.

Publish

• Core + Docker: triggered by this tag
• Vendor shims: publish-packages workflow_dispatch from main

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action is an AI-powered security auditor for Symfony applications that automates the detection of application-level vulnerabilities missed by traditional SAST tools. It targets business logic flaws and multi-file attack chains, providing a comprehensive validation report in console, JSON, SARIF, HTML, or Markdown formats. The action uses adversarial “Attacker” and skeptical “Reviewer” agents to identify and cull false positives over three iterations, ensuring a validated report for security audits in Symfony projects.

What’s Changed

What’s Changed

• ci: add 7-day Dependabot cooldown by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/commit/696801b
• fix(standalone): windows download errors by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/192
• feat(standalone): one liner installation by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/191
• fix(standalone): resolve self-update binary path on macOS by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/194
• test: relax fork/boot slow-test bound by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/195
• feat(command): add scriptable options by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/193
• feat(command): add a doctor preflight command by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/196
• feat(standalone): notify when a newer release is available by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/197
• fix(standalone): throttle failed update checks too by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/199
• fix(standalone): skip update notice after update by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/200
• fix(standalone): honor SSA_NO_UPDATE_CHECK=0 by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/206
• fix(standalone): SSA_INIT fallback under POSIX sh by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/198
• fix(standalone): fail fast on unsupported platforms by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/201
• docs: correct Git Bash install guidance by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/203
• feat(command): add init –force by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/204
• docs: scope PHP/Symfony requirement to the bundle by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/208
• docs: FAQ on auditing older Symfony and PHP by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/209
• docs: correct init options adoption claim by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/207
• fix(command): normalize provider slugs in init by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/202
• fix(standalone): locate only executable binaries by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/205
• chore: release 1.17.0 by @vinceAmstoutz in https://github.com/vinceAmstoutz/symfony-security-auditor/pull/210

Full Changelog: https://github.com/vinceAmstoutz/symfony-security-auditor/compare/1.16.0...1.17.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

RustScript is a tool that interprets and runs Rust scripts without compiling them fully. It allows users to execute Rust code directly from the command line, bypassing the need for Cargo or type checking. The action supports running scripts, validating them without execution, building native binaries, listing supported features, clearing caches, and updating releases.

What’s Changed

Full Changelog: https://github.com/VladasZ/rustscript/compare/v0.2...v0.2.10

• This action is used across all versions by 8 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Legion Runner is an open-source security tool that hardens GitHub Actions against supply chain attacks by monitoring outbound connections, blocking unauthorized destinations, and identifying processes behind them. It provides a comprehensive defense mechanism that runs on pure Node built-ins without dependencies and offers features like egress policy control, process attribution, and file integrity checks.

What’s Changed

Added

• Curated egress presets (allowed-presets): opt-in per-ecosystem allowlists (npm, yarn, pnpm, pip, pypi, cargo, rust, go, maven, gradle, nuget, apt, debian, docker) so block mode “just works” for common toolchains without hand-listing endpoints. e.g. allowed-presets: "cargo, apt". Unit-tested.
• Download integrity verification: the action verifies the legionr-bpf / legionr-fim release binaries against a .sha256 sidecar before running them (the release now attaches the checksums), and fails closed — an unverified/corrupted/tampered download is rejected and the action degrades instead of executing it.
• learned-baseline input (default true): in block mode, also allow destinations previously learned into the Actions cache. Set false to enforce ONLY the explicit allowlist (inline + policy-file + GitHub) with no cache read/write — used by the enforce self-test for deterministic deny.
• File-integrity / tamper detection (Rust legionr-fim agent): snapshots high-value tamper targets at job start (credential/config files, .git config + hooks, and checked-out source) and diffs them at job end, surfacing anything overwritten, deleted, or chmod’d in the summary. Only sha256 hashes are stored — never contents. New inputs file-integrity (auto|off) and fim-extra-paths. file-integrity: auto downloads the agent from the latest release (plain stable Rust, no eBPF toolchain) and degrades to a silent skip if unavailable. Logic lives in legionr-core::fim (unit-tested); the binary is a release asset like legionr-bpf, built + attached by release.yml.
• Package repositories roll-up (📦): the summary now classifies named outbound destinations into their ecosystem/registry (npm, PyPI, crates.io, apt, Docker, Go, NuGet, Maven, Gradle, RubyGems, Alpine, GitHub) and shows a Package repositories reached table — registry, ecosystem, connections, and the process that reached each. Supply-chain risk hides in which registries a build talks to, so we surface them directly instead of leaving you to read IPs. Bare IPs that never got a forward name get a coarse CDN/provider hint (Fastly/Cloudflare/GitHub) via CIDR match — honest about ambiguity (a shared CDN can’t name a registry). Logic in action/repos.js, fully unit-tested.
• Combined cross-job egress report (one summary for the whole run): GitHub has no run-level summary, so each job emits its captured egress as a JSON artifact (node action/report.js emit) and a final egress-report job merges them into a SINGLE table — which job + process reached what — with a package repositories roll-up and a per-job diagnostics block (render). Wired into CI; render is pure and unit-tested. Pairs with job-summary: false so the run shows one combined summary instead of one table per job.
• job-summary input (default true): set false to keep monitoring and enforcement fully active but suppress the connections table in the job summary. Useful when many jobs in one workflow each run the action and you only want the table once (our own CI uses it so a run shows one table, not one per job).
• Secure diagnostics line in the summary: reports which resolution path actually fired (forwarder on/off · captured DNS records N · getaddrinfo route … · named X/Y destinations) so a run that comes back as bare IPs is triagable. Secure by construction — only booleans, counts, and a fixed enum; never the upstream resolver IP, file paths, captured hostnames, or env values.

Fixed

• Block mode no longer hangs the runner at teardown. applyEgressBlock installed a default-deny LEGION_EGRESS chain in OUTPUT and nothing ever removed it, so the runner’s own completion call (to rotating GitHub-backend IPs not in the static seed) was dropped and the job spun until timeout. post() now tears the firewall down (removeEgressBlock).
• Runner hang from leaked daemons. The post step left privileged background processes alive — eBPF agent, DNS forwarder — and the /proc monitor could wedge in a blocking ss subprocess. The monitor now reads /proc/net/tcp directly (no subprocess); daemons are reliably reaped.
• No more spurious “could not resolve” annotations. Allowlist entries that are wildcard parents with no A record of their own (e.g. blob.core.windows.net, actions.githubusercontent.com) used to emit one CI warning annotation each on every run. They are benign: the action skips them, and their subdomains are still observed via PTR / DNS capture (and opened just-in-time in block mode). They are now collected into a single plain-text log line instead.
• Docs/labels: the eBPF mechanism is a tracepoint on sys_enter_connect (not a “kprobe on tcp_connect”); the sampler is /proc-only (the “ss” fallback was removed). Corrected the runtime log line, summary label, and README.
• Outbound connections showed as bare IPs when systemd-resolved owns getaddrinfo. The nsswitch reroute didn’t always stick, so package-repo lookups bypassed the capture forwarder and were never named. The forwarder now targets the real upstream (systemd-resolved’s actual servers, not the 127.0.0.53 stub), and when the bypass is detected but the nsswitch reroute fails, Legion redirects systemd-resolved itself at the forwarder via a resolved.conf.d drop-in — verify-or-revert and restored on teardown, so it never breaks the job’s DNS.
• IPv4-mapped IPv6 destinations rendered as long expanded addresses (0000:0000:0000:0000:0000:ffff:HHHH:HHHH) in the summary. The /proc sampler emitted the expanded form while normalizeIp only collapsed the compressed ::ffff: form. Both now collapse to dotted IPv4 (and the v4/v6 tables dedupe). (Shipped in v1.0.35.)
• Removed dead action/baseline.js; pinned release.yml checkout to v6.

Reliability

• Tests for the paths that kept breaking: the firewall rule builders (egressBlockRules/egressUnblockRules — order, DNS-allow, DROP-last, OUTPUT-jump-removed-first), the checksum parser, the curated presets, and a full-stack PR gate that runs block + DNS-capture + eBPF and asserts the job finalizes (catches any teardown-hang regression), plus the package-repo classifier (host-suffix + CIDR matching). Action test count 19 → 37.

Changed

• Removed em-dashes from the job-summary output (headers, the unresolved-host note, the enforce hint, and empty-cell placeholders) for plainer rendering.
• Name more destinations: route glibc getaddrinfo (curl/apt/cargo/git) through the DNS-capture forwarder via an nsswitch.conf reroute, so hosts resolved by systemd-resolved (which ignores resolv.conf) are now captured and named — not just resolv.conf/c-ares callers. Health-checked and restored on teardown. (A connection to a hard-coded IP with no PTR still shows the IP — there is no name to resolve.)
• More accurate “unresolved destination” note in the summary (a name may have been resolved outside the capture path, vs. a genuine raw-IP connection).
• We dogfood our own action. Every real-work job in this repo (CI, release, eBPF agent, FIM self-test) now runs Legion Runner as its first step (@v1, audit) — our CI is hardened by the product it ships.
• Docs-only PRs skip the build/test matrix (and the release it gates) via paths-ignore; a docs-passthrough workflow reports the same check names green so required checks stay satisfied and README edits stay mergeable without burning CI minutes.
• Our CI now captures names. The dogfooded harden steps run with dns-capture: true (still audit — monitor, don’t firewall), so job summaries show real destination and package-repository names instead of bare IPs, and the capture path is exercised on every run. Safe now that @v1 (>= 1.0.35) carries the teardown + systemd-resolved fixes.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

vibecheck is a tool that automates the detection of AI-generated code smells by using ESLint rules to identify common issues such as hardcoded secrets, empty catch blocks, and SQL injection vulnerabilities. It runs locally without requiring any external dependencies or configuration files. The main purpose of vibecheck is to help developers quickly identify potential problems in their AI-generated codebases, ensuring better security, maintainability, and performance.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

codeguard is a comprehensive tool that automates various repository checks across code quality, design boundaries, security, CI/CD hygiene, AI prompt governance, and repo-specific policy rules. It supports extensive capabilities such as repository exclusions, baselines, waivers, changed-lines diff scans, SARIF output, GitHub annotations, custom rule packs, natural-language custom rules through an optional AI runtime, policy profiles, scan caching, doctor checks, rule discovery from the CLI, native TypeScript/Python quality, design, and security heuristics, and language-specific command checks.

What’s Changed

1.2.1 (2026-07-23)

Bug Fixes

• bound TypeScript semantic analysis to scan corpus (f473e6e)
• bound TypeScript semantic analysis to scan corpus (#61) (685cf99)

• This action is used across all versions by 35 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action compares the versions in package.json and package-lock.json with those of Cargo.toml and pyproject.toml, ensuring they are higher than the latest published tags for each manager. It also provides a way to set custom paths for these files and handle failed checks by setting a fail_on_same_version flag.

What’s Changed

Changes

chore(deps): update npm - all minor and patch updates

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

easySFTP is a fast, secure, and user-friendly GitHub Action for deploying files to an SFTP server. It handles common deploy tasks with minimal configuration and can be expanded for more complex deployments using a config file. The action supports multiple deployment targets, various authentication methods, and advanced features like file skipping, deletion guards, and performance tuning.

What’s Changed

3.0.0 (2026-07-23)

⚠ BREAKING CHANGES

• server was renamed to host, uploads was replaced by source + target, strategy is now mode, ignore is now exclude, host-key-fingerprint is now host-key, config-file is now config.
• Multiple upload mappings require a config file; inline mode supports exactly one deployment.
• All advanced/tuning inputs (concurrency, retries, timeout, stall-timeout, sftp-request-concurrency, sync-fast-path, skip-unchanged, manifest-name, dir-mode, file-mode, preserve-times, max-deletes) moved into the config file.
• All proxy/bastion connection inputs moved into the config file (connection.proxy); only the proxy credentials remain inputs.
• The config file format changed: version: 3, connection settings live in the file, and targets became named deployments.
• Host key verification is required. A run without host-key / known-hosts fails unless allow-any-host-key: true is set explicitly (v2 printed a warning and accepted any key).
• build-mode was removed; the build mode is selected automatically from the action ref (release tags download the verified prebuilt binary, development refs build from source).
• The default log no longer prints one line per file; per-file output moved to log-level: verbose (see Logging changes).
• The delete tombstone input from v1 was finally removed.

Features

• v3 configuration model (inline vs config file, no mixed mode) (#124) (2887530), closes #123

• This action is used across all versions by 349 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates static code analysis for TypeScript and JavaScript projects by identifying unused files, circular dependencies, duplication, complexity hotspots, boundary violations, and design-system styling drift. It provides deterministic findings with typed output contracts and can be integrated into CI workflows to ensure code quality before deployment. The action supports various formats and integrates with popular development tools like LSP servers for language-specific analysis.

What’s Changed

Fallow Impact in your status line

fallow impact statusline gives agent and editor chrome one compact, path-free line with the latest whole-project issue count, its trend, and findings cleared while Impact was tracking. It reads local history only, never runs analysis, migrates data, records telemetry, or prints normal CLI notices.

The Fallow plugin for Claude Code can compose this line with an existing status line. Codex and other agents can use the same stable command through the bundled Fallow skill.

Cloud runtime confidence

Cloud never_called evidence now keeps its provenance. Runtime-observed functions can retain the existing high-confidence deletion recommendation, while inventory-backed, missing, and future provenance remains conservative.

Full Changelog: https://github.com/fallow-rs/fallow/compare/v3.8.1...v3.9.1

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

opseclint is a detection-coverage analyzer that identifies and analyzes the detectability of command-line actions, scripts, or post-exploitation playbooks against various operating systems (Linux/Auditd, Windows/Sysmon, macOS/Endpoint Security) by resolving them to MITRE ATT&CK techniques, telemetry events, and expected detections. It helps developers understand how their actions would be detected if executed in a real-world environment.

What’s Changed

What’s Changed

• Add Dockerfile + GHCR container image by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/10
• Revamp README (Best-README-Template format + shields) by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/11
• Bump version to 0.1.2 by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/12

Full Changelog: https://github.com/Gerrrt/opseclint/compare/v0.1.1...v0.1.2

What’s Changed

• Add Dockerfile + GHCR container image by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/10
• Revamp README (Best-README-Template format + shields) by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/11
• Bump version to 0.1.2 by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/12

Full Changelog: https://github.com/Gerrrt/opseclint/compare/v0.1.1...v0.1.2

What’s Changed

• Add Dockerfile + GHCR container image by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/10
• Revamp README (Best-README-Template format + shields) by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/11
• Bump version to 0.1.2 by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/12

Full Changelog: https://github.com/Gerrrt/opseclint/compare/v0.1.1...v0.1.2

What’s Changed

• Add Dockerfile + GHCR container image by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/10
• Revamp README (Best-README-Template format + shields) by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/11
• Bump version to 0.1.2 by @Gerrrt in https://github.com/Gerrrt/opseclint/pull/12

Full Changelog: https://github.com/Gerrrt/opseclint/compare/v0.1.1...v0.1.2

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

trigger-tree is a local tool that records discovery evidence to improve documentation quality by monitoring which AI coding assistants are accessing project documentation. It helps identify patterns, guardrails, and areas of improvement in the documentation process without relying on cloud analytics or model tokens. The action supports various platforms and provides insights into the health and effectiveness of documentation for teams.

What’s Changed

Both findings below were reported by @BartWaardenburg — thank you.

• The structural inventory follows git’s view of the repository (#16): a file excluded by .gitignore can no longer surface as a router member, orphan, or missing entry point in tt-stats or tt gate. The scope scanner and the structure inventory now share one git-visible source of truth; the filesystem walk remains the fallback outside git repositories.
• Watched surfaces behind a directory symlink are named instead of silently missing (#15): the stats payload gains unfollowed_surfaces (path, target, reason), the health grade states its scope explicitly (evaluable docs), a health driver counts unfollowed surfaces, and tt doctor warns with the paths. Symlinks are still never followed — bytes outside the repository stay out of the deterministic score by design; the boundary is documented in the heat-model guide.

Install (Claude Code)

/plugin marketplace add Hedde/trigger_tree
/plugin install trigger-tree@trigger-tree

Install (Codex, pinned to this release)

codex plugin marketplace add Hedde/trigger_tree --ref v1.24.0
codex plugin add trigger-tree@trigger-tree

Website · Documentation · Changelog · Privacy

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action audits the email-authentication posture of sending domains to ensure they meet best practices, such as SPF, DKIM, and DMARC. It provides a detailed audit report in the job summary and allows users to set severity thresholds to fail builds on regressions. The action is read-only and runs on any runner OS.

What’s Changed

Major correctness + reliability release. The audit’s email-auth verdicts are now governed by a shared conformance spec, verified identically across all three Amino surfaces (this Action, the web tool, and the skill).

Still advisory-by-default — fail-on: advisory continues to never break the build; enforcement is opt-in.

False-pass fixes (a security gate must never call broken config healthy)

• DKIM: an empty p= (revoked selector) and a malformed Ed25519 key no longer read as healthy; real RSA modulus bit-length is read from the key DER (weak = any RSA < 2048).
• DMARC: p= must be none|quarantine|reject (p=banana is no longer “enforced”); tags parsed case-insensitively; RFC 9989 tree walk so a subdomain inherits the org policy instead of a false “No DMARC”; eTLD+1 org-domain (so good.co.uk≠evil.co.uk for report authorization).
• SPF: -ALL handled case-insensitively (no contradictory “no all mechanism”).
• MTA-STS: wildcard matches exactly one label (RFC 8461); enforce requires version: STSv1 + valid mode + integer max_age + ≥1 mx:; policy fetch requires HTTP 200 + text/plain.
• DANE/DNSSEC: DANE requires a DNSSEC-validated (AD-bit) TLSA (RFC 7672); DNSSEC uses the AD bit, which respects the zone cut.

Reliability / new I/O

• New input continue-on-audit-error (default true) and output audit-complete. Empty/invalid domains input now exits non-zero; a transient DNS failure (SERVFAIL/REFUSED) on a critical lookup marks the audit inconclusive (audit-complete=false, passed=false) rather than a silent green — failing the build only when continue-on-audit-error=false.

Engineering

• Zero runtime dependencies preserved. A fixtures-driven conformance runner gates every surface in CI. 69 local tests + 15 shared conformance cases green.

Full plan and per-invariant detail: hireamino/amino-skills#1.

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the installation of the stackit-cli binary on GitHub Actions runners, allowing users to specify a version (latest or a specific semantic version) and cache the installed binary for future use. The action also prepends the cached binary path to the PATH environment variable, making it accessible in subsequent steps of the workflow.

What’s Changed

What’s Changed

🛠️ Dependencies

• chore(deps): update dependencies by @renovate[bot] in https://github.com/jkroepke/setup-stackit-cli/pull/290

Full Changelog: https://github.com/jkroepke/setup-stackit-cli/compare/v1.2.89...v1.2.90

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Codex Review Gate GitHub Action ensures that a PR head has clean Codex review output before passing as a required status check, using Codex’s generative AI capabilities to review pull requests. It maintains a thin workflow in .github/workflows/codex-review-gate.yml and coordinates with GitHub comments, reviews, reactions, and commit statuses for reliable checks.

What’s Changed

What’s changed

• Reconcile official COMMENTED review wrappers with their inline findings instead of treating the wrapper as a second standalone finding.
• Keep incomplete REST/GraphQL snapshots fail-closed until bounded reloads produce a complete, mutually reconciled snapshot.
• Preserve unresolved inline findings as failures while allowing resolved inline findings to stop blocking a later current-head clean result.
• Add regressions for partial snapshots, orphaned parent/child evidence, unknown wrapper grammar, and transient thread retrieval errors.

Compatibility

• Existing action inputs remain compatible.
• completion-signal-buffer-seconds and failed-findings-recovery-mode remain accepted as deprecated legacy inputs.

All release tags (v1.3.1, v1.3, and v1) are signed and resolve to action commit ef41b938a6c57a8a1adde44baaabd29eb8ee4df6.

• This action is used across all versions by 10 repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

NeuroLink is the AI integration platform that provides a single, consistent API for 30+ AI providers and 100+ models. It enables developers to integrate AI into any application with a TypeScript-first approach. Key features include multi-provider failover, intelligent routing, and built-in tools. Users can switch providers with a single parameter change.

What’s Changed

10.5.1 (2026-07-23)

Bug Fixes

• (proxy): retry subscription beta-rejection 400s on the next account (4b67bd1)

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action automates the scan and reporting of security vulnerabilities across AI, cloud infrastructure, and tools. It helps centralize evidence and enforce runtime calls, providing a single Finding with a UnifiedGraph model accessible from CLI, API, UI, and MCP. The action supports scanning from CLI, CI, Docker, or cloud connections on a custom control plane, allowing for comprehensive visibility into security risks across various environments.

What’s Changed

What’s Changed

• fix(deploy): preserve legacy demo overlay before pull by @msaad00 in https://github.com/msaad00/agent-bom/pull/4409
• chore: drop public co-author deny-list and harden release README by @msaad00 in https://github.com/msaad00/agent-bom/pull/4410
• fix(deploy): use shipped driver for postgres migrations by @msaad00 in https://github.com/msaad00/agent-bom/pull/4411
• feat(ui): make context and investigation graphs the canvas hero by @msaad00 in https://github.com/msaad00/agent-bom/pull/4412
• feat(ui): punchier graph captures and skill capability contracts by @msaad00 in https://github.com/msaad00/agent-bom/pull/4414
• chore(ops): fail closed on missing UI Hub tag and fix pilot compose by @msaad00 in https://github.com/msaad00/agent-bom/pull/4413
• feat(ui): distinct graph proof shots and labeled relationships by @msaad00 in https://github.com/msaad00/agent-bom/pull/4415
• chore(deps): bump aws-actions/configure-aws-credentials from 6.2.2 to 6.2.3 by @dependabot[bot] in https://github.com/msaad00/agent-bom/pull/4416
• chore(deps): bump lucide-react from 1.25.0 to 1.26.0 in /ui by @dependabot[bot] in https://github.com/msaad00/agent-bom/pull/4417
• fix(deps): bump postcss to 8.5.12 for CVE-2026-45623 by @msaad00 in https://github.com/msaad00/agent-bom/pull/4420
• fix(deploy): harden EKS reference install path by @msaad00 in https://github.com/msaad00/agent-bom/pull/4421
• docs(arch): Python-primary runtime; optional Go sidecar later by @msaad00 in https://github.com/msaad00/agent-bom/pull/4423
• chore(release): 0.97.5 by @msaad00 in https://github.com/msaad00/agent-bom/pull/4422

Full Changelog: https://github.com/msaad00/agent-bom/compare/v0.97.4...v0.97.5

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Lambda MicroVM GitHub Runner is an action designed to run GitHub Actions jobs on AWS Lambda using MicroVMs, providing a way to extend Lambda’s 8-hour execution time by utilizing the lightweight nature of MicroVMs. This tool automates the setup and teardown of AWS resources required for running these actions, including creating IAM users, roles, and logs. It also supports configuring containerized workflows inside the runner, making it suitable for multi-container applications.

What’s Changed

What’s Changed

• Add suspended warm server pools by @neebs12 in https://github.com/neebs12/lambda-microvm-github-runner/pull/14

Full Changelog: https://github.com/neebs12/lambda-microvm-github-runner/compare/v1...v1.1

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action retrieves secrets from Project Vault and exports them as masked environment variables in a workflow. It automates the retrieval of secrets scoped to specific projects using machine-user API keys, ensuring that each secrets input refers to the same project, and provides options for handling errors gracefully. The action does not implement its own HTTP client or token exchange logic.

What’s Changed

Retrieve secrets from Project Vault and export them as masked environment variables in your GitHub Actions workflow — no custom HTTP calls, no plaintext secrets in logs.

Usage

- uses: nestormata/vault-action@v1
  with:
    vault-url: https://vault.example.com
    api-key: ${{ secrets.VAULT_API_KEY }}
    secrets: |
      <PROJECT_ID>/<CREDENTIAL_NAME> as <ENV_VAR_NAME>

See the README for full setup (machine-user + API key creation), multi-secret syntax, continue-on-error semantics, and SHA-pinning guidance.

Highlights

• Machine-user API key auth, scoped to one project per step
• Automatic masking of every retrieved value
• 10-second network timeout per credential, so an unreachable vault fails fast instead of stalling your job
• continue-on-error input to soften only “vault unreachable” failures — application-level errors (bad key, missing credential, wrong scope) always fail the step
• Runs on Node 24

• This action is used across all versions by 10 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automatically pulls new versions of Go modules when tags are created, ensuring that your module is immediately available and up-to-date on platforms like pkg.go.dev. It supports both standard and submodule version tags and allows customization of proxy configuration, import paths, and Go versions. The action is triggered by release events and uses the actions/setup-go toolchain for setup.

What’s Changed

1.1.31 (2026-07-23)

• This action is used across all versions by 3 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action converts Notion pages to Jekyll markdown files, extracting properties such as category and tags. It also downloads block assets like images and files from Notion, ensuring SEO compatibility by adding necessary front matter to posts. The action is designed to be run in a GitHub Actions workflow and can be customized via environment variables for different configurations.

What’s Changed

What’s Changed

• date range support by @pirafrank in https://github.com/pirafrank/notion-to-jekyll/pull/2
• dry run
• log improvements
• enhancements and bug fixes

Full Changelog: https://github.com/pirafrank/notion-to-jekyll/compare/v1...v2

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of setting up a new API repository by bootstrapping a Postman workspace, uploading an OpenAPI specification, generating test collections, and syncing them with the repository. It also creates CI workflows to rerun tests on every push, pull request, and schedule. The action is designed to provide comprehensive, executable testing solutions for APIs in GitHub repositories.

What’s Changed

What’s Changed

• Release v2.1.8 by @jaredboynton in https://github.com/postman-cs/postman-api-onboarding-action/pull/90

Full Changelog: https://github.com/postman-cs/postman-api-onboarding-action/compare/v2.1.7...v2.1.8

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the creation of a Postman workspace from an OpenAPI specification, generating baseline, smoke, and contract collections. It also includes dynamic contract tests that cover various protocols such as gRPC, SOAP, GraphQL, AsyncAPI, and MCP. The action is part of the broader Postman API Onboarding suite and requires either a public HTTPS URL or a file path to the spec in the repository.

What’s Changed

What’s Changed

• Hydrate org reconciliation markers from collection export by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/140
• Release v2.10.19 by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/141

Full Changelog: https://github.com/postman-cs/postman-bootstrap-action/compare/v2...v2.10.19

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action, Postman Onboarding: GCP Spec Discovery, automates the discovery and export of OpenAPI specifications from Google Cloud services such as GCP APIs and Apigee proxies. It can automatically resolve a specification based on repository context or labels if available, or explicitly specify an API ID for a known configuration. The action uses Application Default Credentials (ADC) or Workload Identity Federation for authentication and supports exporting every candidate specification to facilitate Postman onboarding processes.

What’s Changed

What’s Changed

• fix(test): select npm CLI on Windows by @jaredboynton in https://github.com/postman-cs/postman-gcp-spec-discovery-action/pull/24
• chore: prepare v1.1.8 release by @jaredboynton in https://github.com/postman-cs/postman-gcp-spec-discovery-action/pull/25

Full Changelog: https://github.com/postman-cs/postman-gcp-spec-discovery-action/compare/v1.1...v1.1.8

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action automates the deployment, testing, and quality analysis of Salesforce changes using the @sfdt/cli tool. It provides features like interactive deployment workflows, automated release manifest generation, parallel Apex test execution, code quality analysis with AI-powered fix plans, pre-release validation checklist, and multi-package project support.

What’s Changed

What’s Changed

• chore: promote develop → main (extension v0.8.1) by @scoobydrew83 in https://github.com/scoobydrew83/sfdt/pull/268

Full Changelog: https://github.com/scoobydrew83/sfdt/compare/v0.18.2...ext-v0.8.1

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automatically translates i18n locale files in a repository using the Shipi18n platform. It supports multiple languages, placeholder preservation, and incremental translation, automating the process of maintaining internationalized content. The action also generates pull requests for review and self-corrects failed translations.

What’s Changed

Relicensed to Apache-2.0.

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The bench-action GitHub Action automates benchmarking of TypeScript files using the @soroush.tech/bench package in a Docker sandbox. It fails the CI build if any case drops below a specified minimum speed ratio compared to the baseline case, and posts results as a sticky PR comment using the bench bot if configured with the id-token: write permission. The action is easy to use as it doesn’t require installation or setup beyond selecting a release tag.

What’s Changed

bench-action@v1

First public release of Bench Gate — a performance gate for pull requests.

It runs your repo’s @soroush.tech/bench *.bench.ts suites inside a CPU/RAM-pinned Docker sandbox, compares every case against a baseline case, fails the job when any case drops below your minimum speed ratio, and posts the results as a single sticky PR comment that updates in place on every push.

One uses: step, zero setup.

Highlights

• Pinned sandbox, stable numbers
  Every suite runs in a Docker container pinned to a single CPU (--cpuset-cpus) with a hard CPU quota and a memory cap (swap disabled), so timings stay comparable run-to-run even on shared runners. rounds: N gates on the median of N runs for extra noise immunity.

• Ratio gate
  Pick a baseline-case and a min-ratio (percent). Any case slower than the threshold fails its file; any failed file fails the job. Gate against your last published version by declaring it via npm’s latest dist-tag directly inside the bench file.

• Zero-config results comment
  github-token defaults to the workflow’s automatic runtime token—no secrets to wire. Pass '' to skip commenting entirely; the gate verdict is never affected by comment delivery.

• Branded comments (optional)
  Install the bench GitHub App and grant id-token: write. The comment is authored by the bench bot via an OIDC-verified relay, with automatic fallback to a direct GitHub comment if the relay is unavailable.

• Everything baked in
  The bundled runner and vendored sandbox ship with the release tag. GitHub-hosted runners already provide Docker, so there’s nothing to install, build, or download in your workflow.

Getting Started

jobs:
  bench:
    runs-on: ubuntu-latest
    permissions:
      contents: read
      pull-requests: write # Required for the results comment
      id-token: write # Optional, enables branded "bench bot" comments

    steps:
      - uses: actions/checkout@v5

      - uses: soroush-tech/bench-action@v1
        with:
          bench-dir: bench
          baseline-case: previous
          min-ratio: "80"
          rounds: "5"

Branded Comments

Branded comments are a one-time, secret-free setup:

1. Install the bench GitHub App on your repository.
2. Keep id-token: write in your workflow permissions.

If you skip both, everything still works—the results comment is simply posted by github-actions[bot].

Requirements

• A Linux runner with Docker (all GitHub-hosted Linux runners qualify).
• The Node 24 Actions runtime. Any current GitHub-hosted runner works; self-hosted runners must be recent enough to include Node 24.
• Pin a release (@v1 or a version tag). @main does not include build artifacts and will not run.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Sigbound automates the process of merging code changes from multiple AI coding agents in parallel, resolving conflicts using a model and verifying merges before they land. It uses Git worktrees to handle each agent’s work independently and combines non-conflicting changes efficiently. The action is customizable with various commands for planning, running agents, resolving conflicts, verifying merges, and repairing failed builds, allowing users to harness their own AI models for effective collaboration.

What’s Changed

Changelog

• cbfd3b7ac0ba5d0f1689c06599f2a6cc243c898b docs: correct /ui auth prose to match the SPA-shell pattern
• 5cbcf58af0c98e3e91bb2bcb5301d35c0815a6cd fix: close bare ‘’ fail-open in -env- allowlists (issue #56)
• b41f5a535d184afcfb92118ee41f9af2e6f4a53e fix: don’t count -verify-cache hits toward verify.invocations/wallMs
• e6043c25b71cf264512d597781f4bfbec72cb8d8 fix: resolve singleton group heads to OIDs in IntegrateOCC
• b894a53137a70bdeba7c22d4ea81b23e9db838c3 fix: serve /ui unauthenticated so the token field is reachable

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

RustScript is a tool that allows running Rust scripts without compiling them. It supports a practical subset of the language and provides features like rust check to validate scripts without executing them. The interpreter does not implement a second type system, but it handles errors and panics as expected for compiled Rust code.

What’s Changed

Full Changelog: https://github.com/VladasZ/rustscript/compare/v0.2...v0.2.7

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

cowork-harness is a scriptable, CI-friendly test harness that accurately reproduces Claude Cowork’s observable runtime contract. It supports various testing scenarios and environments, including local skills and live tiers with multi-node setups and Docker containers. The tool helps developers automate their skill testing across different platforms and ensures consistent behavior with real Cowork runs.

What’s Changed

Added

• check:versions closes the stale-pin gap a docs audit found. Three new enforcement surfaces: every companion-skill references/*.md (baseline desktop-X.Y.Z) pin must match SKILL.md’s tracks-harness baseline; task-recipes.md now carries the same guarded Tracks `cowork-harness X.Y.Z` stamp as the other references (and is a bump-version target); DESIGN.md’s one present-tense “currently X, per baselines/desktop-Y.json” sentence must name the max committed baseline + its agentVersion (its dated verification-pass notes stay exempt). A new docs-index sync test keeps README’s Documentation table and docs/README.md’s Guides table from drifting apart (that drift is how critique.md went missing from the README table).

• doctor warns when a pulled agent image is behind the published one. For the container/hostloop/cowork tiers, a new advisory image-freshness check compares the local agent image’s registry digest against the current published ghcr.io/yaniv-golan/cowork-agent-base:2 (or -full) and warns — with a re-pull + retag remedy — when they diverge. Best-effort and never blocking: a locally-built image (no registry digest), an offline host, docker buildx being unavailable, or a custom COWORK_AGENT_IMAGE is a quiet skip, never a false “stale”; only a pulled image incurs a network probe. The doctor --output-format json envelope gains an image-freshness entry (the open checks[].id set — SPEC §12 — already permits this).

• critique now stamps verdictProvenance on every report (JSON key + text “verdict scope” line): the verdict is an advisory self-run, not an independent attestation.

• critique --skill <name> — plugin-aware grading. A multi-skill plugin root (skills/<name>/SKILL.md, no root SKILL.md) previously graded a missing file, downgrading every coverage finding to “not adjudicable”. --skill selects the invoked skill’s folder for the packager (selection only — the positional folder is still what both turns mount, and fingerprint.skillHash is unchanged: per-plugin, not per-skill); a multi-skill root with no --skill is refused before any model spend, naming the available skills; a single-skill plugin auto-selects with a notice. The evidence package now also carries the invoked skill’s agents/<name>.md body and bounded references/*.md content (not just filenames), and the report carries an advisory skillInvocationObserved (false = the graded run’s own skillActivity never mentions the selected skill).

• critique persists run-dir artifacts. Every critique writes critique-report.json (the machine-readable report); when the evaluator ran, critique-evidence-package.txt (the exact armored corpus it graded against — re-grade a disputed finding offline); on an instrument failure (exit 2), critique-salvage.json (the self-report + each evaluator pass’s RAW reply captured pre-parse). New --out <path> also writes the selected-format report to a file. The two not-built limitations these close (evidence-not-persisted, report-stdout-only) are removed from the registry.

• Per-critique cost across all four workloads. The report’s costUsd sums the task turn, reflection turn, and BOTH evaluator passes (whose usage the transport previously discarded), marked INCOMPLETE whenever any workload is unpriced. The header also carries the pinned fidelity plus the graded turn’s recorded gradedEffectiveFidelity/gradedBaseline, and the graded run’s resolved gate answers are echoed as copy-pasteable --answer lines (JSON: gateAnswers).

• critique evaluator parse is per-item tolerant. One malformed item in an {"items":[...]} reply previously discarded the whole document (“no valid JSON found”, a broken discovery run). Valid items now survive; malformed ones are dropped AND counted (droppedEvaluatorItems, surfaced in both report formats). The integrity canary is recognized by its idea alone (a mutated echo still proves instruction-following) and stripped before dedup, so a full-document + canary-only-restatement reply no longer trips the ambiguity throw. Fail-loud preserved: garbage, or all-malformed-with-no-canary, still throws — now naming which field check failed.

• Sub-agent research is observable end-to-end. A sub-agent’s WebSearch never enters the main toolCounts/webSearches[]; its query + result are now captured from the child session transcript as subagents[].webSearches (bounded, live/record lane only — absent on replay, and absence is never evidence of no research), surfaced by the new trace --view subagent-research, and packaged into critique’s evidence as a “Sub-agent research” section so “researched” claims are groundable.

• Each critique item now carries a findingFingerprint (sha16 over the normalized idea + classification + recommendedAction, deliberately excluding the input-specific evidence excerpt) — clusters the SAME finding across DIFFERENT inputs, complementing skillHash (same skill across fixes). docs/critique.md gains a “Reproduction” section documenting the ≥2-run discipline.

• SKILL.md truncation is now reported distinctly: a readable-but-oversized SKILL.md marks the report skillMdTruncated (“the evaluator graded a cut copy”) instead of being indistinguishable from a fully packaged one; only missing/unreadable still forces the mechanical “already-covered” downgrade — the --help/docs limitation wording now says so.

Changed

• Published agent images (cowork-agent-base / cowork-agent-full) now carry OCI metadata labels (org.opencontainers.image.{title,description,source,documentation,licenses}), so their GHCR package pages render a description, repo link, license, and the “contains no Anthropic binary” provenance note instead of appearing bare. The publish-image workflow sets the full variant’s title/description via --label so its page isn’t mislabeled by the base image’s baked-in defaults.
• critique evidence caps raised: SKILL.md 16KB→64KB, transcript 16KB→32KB, overall package 48KB→144KB (the overall cap sits above the worst-case per-section sum, agents/references/research sections included), so a flagship-sized (~51KB) SKILL.md no longer grades permanently truncated. Increases per-critique evaluator token cost on large skills (~2–2.5×).
• sync’s two code-tripwire warnings (the getMcpSkillSources caller-count and MCP-skills-capability checks) now carry a self-contained instruction — re-verify whether MCP servers can contribute skills and whether the harness must model MCP-contributed skill sources — instead of pointing at a reference the published repo does not carry.
• Pinned sessions stamp their fidelity tier on session.json, and a cross-tier --resume fails loud (pre-spawn, with a “re-run at --fidelity <stamped>” remedy). The agent’s native conversation store is tier-LOCAL — container persists it under the work tree, hostloop under the host config dir — so resuming a session at a different tier would hand the binary a --resume for a conversation its store has never seen. Legacy stampless manifests (pre-dating the stamp) are let through with a warning; every manifest written from now on carries the stamp. readSessionManifest gains a required expectedFidelity argument.

Docs

• README documents that the agent images are published to GHCR and can be pulled + retagged instead of built from scratch (:2 floats to the latest release; :2-<version> pins an immutable per-release build), and that the harness resolves the local tag — so a stale local image shadows the published one; re-pull after upgrading. The doctor command row notes the new freshness warning.

Fixed

• hostloop uploads bullet advertised a non-readable path. The dynamic “## Shell access” section rendered the uploads mapping’s file-tool side as dirname(upload.hostPath) — the user’s original source dir, which the path-containment gate does not allow — while the base prompt pointed at the correct staged dir. An agent following the bullet got “outside this session’s connected folders” and worked around it via copy-into-outputs + rm, tripping a spurious outputs-delete fail. The bullet now advertises the staged uploads dir — the SAME hoisted value the path gate allows, so the prompt and the gate cannot disagree. (Live-verified: hostloop --upload + Read-tool read succeeds at the advertised path.)
• The task-turn timeout kill message now names --timeout and the 10-minute default; a “missing” SKILL.md report note now points at --skill / the invoked skill’s folder (the multi-skill-plugin-root cause) instead of only stating the symptom.
• Shipped-doc corrections that induced field misdiagnoses: the cowork-harness skill’s gotcha #8 now states that production enforces outputs delete-deny (EPERM + approval) and the harness gap is detection-only; docs/critique.md documents that a WebSearch produces no search-host entries in the container egress.log (an api.anthropic.com-only log is consistent with research working — live-verified with a first-party capture) and that sub-agent searches don’t increment the main toolCounts.WebSearch; the shipped fidelity reference explains the hostloop Read-vs-bash path split including uploads readability.
• critique now runs at --fidelity hostloop as well as container — the container-tier pin is lifted. The pin existed because the reflection turn resumes the task turn and resume-continuity was only proven for the container Linux ELF; a live two-turn proof at hostloop’s native agent binary (test/live-contract.test.ts, “resume-continuity proof at hostloop”) cleared it, and a live critique --fidelity hostloop e2e validates the full protocol there. microvm/protocol/cowork stay refused, each with its own stated reason: the single container-tier-only limitation is replaced by three tagged ones — microvm-tier-refused [unverified] (a resume-continuity proof at the microVM guest would lift it), protocol-tier-refused [not-built] (protocol never plumbs a session id/--resume), and cowork-tier-refused [deliberate] (the synced loop gate would make the graded tier baseline-dependent, adding noise to skillHash-paired generation comparisons).
• skill accepts --allow-host-writes (and critique forwards it to BOTH turns) — the hostloop writable-connected-folder consent that previously only chat (its own flag) and run (the allow_host_writes: true scenario field) could grant; a plain skill --fidelity hostloop --folder X was refused with no way to consent at all. Folder-less runs (skill dir + uploads) still need no consent — uploads and skill/plugin mounts are read-only.
• critique’s spaced flag parser no longer silently grabs the next flag as a value. critique <folder> --prompt --output-format json (a forgotten --prompt value) swallowed --output-format as the prompt AND dropped the real flag, then ran a four-workload critique on the wrong input. The spaced form now fails loud and points at the equals escape hatch (--prompt=<value>) for a value that intentionally starts with -.
• critique’s subprocess byte cap is now one combined stdout+stderr budget. boundedSpawn kept two independent per-stream counters, so a looping or hostile child splitting output across both streams could buffer ~2× the documented cap before either tripped. It now charges both streams against a single budget and slices the terminal chunk to the remaining room, so captured output never exceeds the cap.
• critique’s verbatim citation-grounding is now case-sensitive. The mechanical check that drops any finding whose cited evidence is not a verbatim excerpt normalized whitespace AND folded case — so a case-altered (paraphrased) citation resolved as “grounded,” weakening the principal defense against evaluator hallucination. It now matches case-exact; whitespace reflow stays tolerated (models reflow spacing when quoting — they don’t change case).
• critique’s “Attached inputs” evidence no longer reports (none) when the uploads dir is unreadable. listAttachedInputs caught every readdirSync failure identically, so an unreadable uploads dir (EACCES/ENOTDIR/…) collapsed to the same (none) as a legitimately-absent one — telling the evaluator “the agent correctly saw no file” when attachment presence was actually UNKNOWN. It now distinguishes ENOENT (absent) from a genuine read fault and surfaces the uncertainty loudly.
• critique flags a partly-corrupt archived turn-1 transcript instead of grading it as clean. readTurn1Transcript skipped malformed JSONL rows (resilient) but returned degraded: false on the first valid transcript record regardless — so archive corruption vanished from evidence health and the evaluator graded a silently-incomplete transcript as solid ground truth. It now scans the whole (turn-1-only) archive and sets turn1SliceDegraded when a malformed row was skipped or the archive doesn’t hold exactly one transcript record; the transcript is still delivered, just flagged.
• critique flags a crashed task turn instead of grading evidence from it. taskTurnInfraFailure only treated a killed (timed-out / byte-capped) task turn as an instrument failure — a task that exited nonzero without ever printing a parseable result envelope (a crash after the early [status] line) slipped through, and its [status]-recovered run dir was reflected on and graded as a legitimate task. It now also flags a nonzero exit with no parseable envelope. Deliberately narrow: a completed run that reported a failing verdict (ok:false / result:"error" with a valid envelope) stays gradeable.
• A nested unreadable output subtree no longer persists a partial file list as complete. The workspaceFiles / artifacts list was recorded as UNAVAILABLE (undefined, the evidence-unavailable convention) only when the workspace ROOT was unobservable; a nested EACCES/EIO subtree left the walk partial (walkComplete: false) but still persisted the incompletely-enumerated list as if complete. An authored file inside the unreadable subtree then vanished with no signal, so an absence-sensitive consumer (delivered_clean / ended_with_question, the replay diff, scaffold, file_exists) could read it as absent — a silent false-clean. All three RunResult producers (run, partial-salvage, chat) now route through one shared trustedWorkspaceFiles gate that collapses a missing root OR any incomplete walk to undefined, and the run lanes emit a ::warning:: naming the unreadable subtree.

What’s Changed

• fix(security): fast-uri HIGH advisories + CodeQL URL-substring alert by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/64
• chore(deps): Bump the actions group across 1 directory with 3 updates by @dependabot[bot] in https://github.com/yaniv-golan/cowork-harness/pull/63
• feat(doctor): warn on stale pulled agent image; label GHCR images by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/65
• fix(critique): reject a flag-looking value in the spaced form (no silent positional-grab) by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/66
• fix(critique): one combined stdout+stderr byte cap in boundedSpawn (was 2 independent caps) by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/67
• fix(critique): make verbatim citation-grounding case-sensitive (was case-folded) by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/68
• fix(critique): don’t report “(none)” attachments when the uploads dir is unreadable by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/69
• fix(critique): degrade a partly-corrupt archived turn-1 transcript (was silently clean) by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/70
• fix(critique): flag a crashed task turn instead of grading evidence from it by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/71
• fix(run): record workspaceFiles/artifacts unavailable on an incomplete workspace walk (#54) by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/72
• release: 1.8.0 by @yaniv-golan in https://github.com/yaniv-golan/cowork-harness/pull/73

Full Changelog: https://github.com/yaniv-golan/cowork-harness/compare/v1...v1.8.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Agent Lint is a linter tool that validates configuration files and directories related to Claude, Cursor, Codex, and MCP agents. It checks for various rules such as plugin paths, hook paths, skill frontmatter, agent fields, prompt quality, and more. The action can be used via GitHub Actions or pre-commit hooks, with the option to run specific rules or suppress them based on a configuration file.

What’s Changed

What’s Changed

• Fix manual upgrade handoff by @zhupanov in https://github.com/zhupanov/agent-lint/pull/656
• Fix false positives from SRE pedantic scan by @zhupanov in https://github.com/zhupanov/agent-lint/pull/658
• Release v4.0.3 by @zhupanov in https://github.com/zhupanov/agent-lint/pull/659

Full Changelog: https://github.com/zhupanov/agent-lint/compare/v4...v4.0.3

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action, zizmor-action, automates the execution of the zizmor security scanning tool in a GitHub workflow. It provides detailed analysis of code to identify potential vulnerabilities and helps developers triage issues more efficiently. The action supports both public and private repositories with various configurations like severity thresholds, token management, and output formats. Users can integrate it into their workflows for continuous security monitoring without relying on advanced security features.

What’s Changed

zizmor 1.28.0 is now the default version used by the action.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

slopscore is a transparent linter that measures the density of formulaic, generic, low-specificity, and over-polished writing patterns in text. It reports per-dimension scores and evidence spans to help identify specific writing issues, encouraging clearer, more specific prose. The tool is designed as a prose linter rather than an AI detector for authorship, focusing on detecting common writing patterns that are characteristic of AI-generated or low-effort content.

What’s Changed

Adds a structural-metaphor tell to insight_signaling — a class a human reader caught that both linters (and a curated grep) missed.

What’s new

insight_signaling now flags anatomical/structural metaphors for an organizing principle:

• Noun-of-noun (“the spine of the argument”, “the backbone of the piece”) — INSIGHT_STRUCTURAL_METAPHOR.
• Copular (“The spine is the same evidence…”) — INSIGHT_STRUCTURAL_COPULA.

Both are context-gated: the noun-of-noun form requires a prose/argument noun, and the copular form gates its complement to abstract nouns — so literal senses (lumbar spine, backbone network, construction/React scaffolding, book spine, “a column of vertebrae”, “a fiber-optic link”) do not fire. Bare “through-line” (legitimate in film/writing) is --broad-only.

Ships with literal-negative unit tests (the real guard for this rule class). The JSON report shape is unchanged; SCHEMA_VERSION stays 0.8.0.

Full changelog: https://github.com/jman4162/slopscore/blob/main/CHANGELOG.md

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the bulk configuration of various GitHub organization settings across multiple organizations using a declarative YAML configuration file. Key features include syncing custom property definitions, values, rulesets, issue types and fields, member privileges, repository policies, and .github repository files. The action supports dry-run mode for previewing changes and provides per-organization overrides via YAML configuration. It also offers rich job summaries with per-organization status tables for easy tracking of changes. The action is designed to support GitHub.com, GHES, and GHEC platforms.

What’s Changed

What’s Changed

• feat: add custom property value sync by @joshjohanning in https://github.com/joshjohanning/bulk-github-org-settings-sync-action/pull/72

Full Changelog: https://github.com/joshjohanning/bulk-github-org-settings-sync-action/compare/v1.12.4...v1.13.0

• This action is used across all versions by 3,357 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of generating Python data models from various schema definitions, including OpenAPI 3, AsyncAPI, JSON Schema, Apache Avro, XML Schema, Protocol Buffers/gRPC, GraphQL, and MCP tool schemas. It supports converting raw data (JSON/YAML/CSV) into Python model output types, retargeting existing Pydantic, dataclass, or TypedDict classes, and outputs models in different styles like Pydantic v2, Pydantic v2 dataclass, dataclasses, TypedDict, or msgspec. The action handles complex schemas with $ref, allOf, oneOf, anyOf, enums, and nested types, producing type-safe, validated code that is ready for IDEs and type checkers.

What’s Changed

Breaking Changes

• Preserved additionalProperties value constraints change generated output - JSON Schema additionalProperties (and propertyNames/mapping) value schemas that carry constraints (e.g. minimum/maximum, minLength/maxLength/pattern, array item constraints, allOf-merged primitives) now retain those constraints in the generated model instead of dropping them. This produces new TypeAlias/RootModel/TypeAliasType definitions and changes dict value type hints, so output differs for existing schemas. For example, a mapping value that previously generated dict[str, Literal['fixed']] now generates a dedicated alias preserving the constraints (#3616):

# Before
class DictModel(RootModel[dict[str, Literal['fixed']]]):
    root: dict[str, Literal['fixed']]

# After
DictModelAdditionalProperty = TypeAliasType(
    "DictModelAdditionalProperty",
    Annotated[Literal['fixed'], Field(max_length=100, min_length=1)],
)


class DictModel(RootModel[dict[str, DictModelAdditionalProperty]]):
    root: dict[str, DictModelAdditionalProperty]

What’s Changed

• Update CHANGELOG for 0.69.0 by @dcg-generated-docs[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3613
• Update release benchmark data by @dcg-generated-docs[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3614
• [pre-commit.ci] pre-commit autoupdate by @pre-commit-ci[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3617
• Preserve additionalProperties value constraints by @chuenchen309 in https://github.com/koxudaxi/datamodel-code-generator/pull/3616
• Rename a msgspec field named “field” to avoid shadowing the field import by @chuenchen309 in https://github.com/koxudaxi/datamodel-code-generator/pull/3615
• Bump the github-actions group with 8 updates by @dependabot[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3618
• Update types-setuptools requirement from <70,>=67.6.0.5 to >=67.6.0.5,<84 by @dependabot[bot] in https://github.com/koxudaxi/datamodel-code-generator/pull/3584

Full Changelog: https://github.com/koxudaxi/datamodel-code-generator/compare/0.69.0...0.70.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The agent-test-verifier (atv) is a static analysis tool that detects patterns commonly used by coding agents to bypass bug fixes and fool test suite verification. It scans diff files for changes that could lead to fake passing tests, such as force-pass hooks or assertion weakening. The tool can be used in CI pipelines to ensure that code changes actually fix bugs rather than just manipulate the test output.

What’s Changed

First public release of the agent-test-verifier GitHub Action: a static, deterministic check that catches the patterns coding agents use to fake a passing test suite — force-pass conftest.py hooks, neutered/trivialized assertions, newly skipped tests, and null tests that cannot fail.

Use it

name: agent-test-verifier
on: pull_request
jobs:
  atv:
    runs-on: ubuntu-latest
    steps:
      - uses: actions/checkout@v4
        with:
          fetch-depth: 0
      - uses: LaterKidsXD/agent-test-verifier@v1

Findings show up as inline annotations on the PR diff, a findings table on the run’s Summary page, and a red check when anything meets the fail-on bar. See the permanently open demo PR with every pattern planted.

1.0.1

• Marketplace branding metadata in action.yml
• README: live demo PR link
• Fixed atv.__version__ drift (0.1.0 → in lockstep with the package version)

1.0.0

• --format {text,json,github,markdown} CLI (--json kept as alias)
• Composite action: base-ref resolution, severity-gated annotations, step-summary table, honest exit codes (0 clean / 1 findings / 2 usage error — never a silent false “clean”)
• Self-check workflow dogfooding the action on this repo’s own PRs

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action takes vulnerabilities from Trivy scans and creates Jira issues for each vulnerability. It automatically adds these issues as subtasks under a parent issue and can move them between states based on whether they have a fix version or not. The action also ensures that CVEs are shared across multiple projects in the same Jira instance, with project-specific versions marked on each issue.

What’s Changed

• Merge pull request #17 from levigo/fix/NF-3145_CVE_Score (c4ee83e)
• fix(NF-3145): Implement score selection analog with Trivys severity filter. (ce50175)
• Merge pull request #16 from levigo/fix/NF-2994-link-to-cve-database-in-generated-cve-tickets-doesnt-work-404-error (81c9666)
• fix(NF-2994): update URL resolution to return multiple references for CVE tickets (e3d857b)
• Merge pull request #14 from levigo/fix/NF-2994-link-to-cve-database-in-generated-cve-tickets-doesnt-work-404-error (ada9e08)
• fix(NF-2994): update URL resolution to return multiple references for CVE tickets (1cb2f2d)
• fix(NF-2994): update URL resolution to return multiple references for CVE tickets (1e8feb2)
• Merge pull request #13 from levigo/feature/esm (6130050)
• fix(INF-346): json syntax error (2b9e63b)
• Merge pull request #12 from levigo/feature/esm (a3852cf)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary: lgtmaybe is a provider-agnostic GitHub Action that automatically reviews pull requests by analyzing the diff. It provides a comprehensive review including logic and correctness bugs, security vulnerabilities, missing tests, outdated code, performance regressions, unnecessary complexity, intent misalignment, and potential ponytail (code redundancy) issues. The tool uses OpenAI models to generate inline comments and a summary of the PR’s changes, handling sensitive information securely and efficiently.

What’s Changed

0.12.2 (2026-07-22)

Bug Fixes

• homebrew: prefer wheels so litellm >=1.92 sdist never builds in the sandbox (#189) (df6e7e1)

Dependencies

• bump the python-dependencies group with 6 updates (#187) (5615a79)

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Totem is a tool that automates the process of creating and enforcing coding rules within repositories using Markdown lessons. It ensures that project-specific knowledge and lessons are stored alongside the code, preventing inconsistencies between sessions and reducing the need for constant re-explanations. By compiling lint rules from these lessons, Totem provides a deterministic, offline linter that helps maintain architectural integrity and velocity.

What’s Changed

Cohort-link bump (no direct package changes). See .changeset/config.json for the fixed-cohort definition.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action installs the iso8583tool CLI, a tool for generating and parsing ISO 8583 messages, on your CI/CD pipelines. It downloads prebuilt binaries instead of building from source, ensuring fast execution times across Linux, macOS, and Windows platforms. The action allows you to specify a version to install and verify checksums or attestation.

What’s Changed

Full Changelog: https://github.com/nao1215/setup-iso8583tool/commits/v0.1.0

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action linkml-scala-action automates the validation and generation of LinkML schemas using Node.js. It supports various commands like validate and generate, with options to specify schema files, strict mode, and output directories. The action can emit annotations to GitHub PRs, making it easier to track issues in schema definitions.

What’s Changed

Tracks linkml-scala v0.11.0.

Bundles @neverblink/linkml@0.11.0.

• Pin uses: NeverBlink-OSS/linkml-scala-action@v0.11.0 for reproducibility.
• Pin @v1 for automatic patch/minor updates.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

AI Harness Doctor automates the auditing of AI agent configurations across repositories to ensure consistency and accuracy in instruction sets. It helps consolidate scattered guidance into a single AGENTS.md file, improves agent answers by consolidating conflicting instructions, and measures improvements in latency and cost. The tool also detects issues such as overlapping files, mismatched declarations, and conflicts within the same scope.

What’s Changed

What’s Changed

• feat(registry): recognize Warp, Firebase Studio, Goose, and Kiro agents by @NieZhuZhu in https://github.com/NieZhuZhu/ai-harness-doctor/pull/325
• fix(scan): stop reading “npm provenance” release prose as npm package-manager usage by @NieZhuZhu in https://github.com/NieZhuZhu/ai-harness-doctor/pull/326

Full Changelog: https://github.com/NieZhuZhu/ai-harness-doctor/compare/v1.15.2...v1.16.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the provisioning of a Postman workspace by generating baseline, smoke, and contract collections based on an OpenAPI specification. It includes dynamic contract tests that validate request/response schemas, security checks, and multi-protocol support. The action provides executable contract tests and logs the Postman CLI in for spec linting.

What’s Changed

What’s Changed

• fix: converge concurrent generation ownership by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/110
• fix(release): publish verified bootstrap artifacts by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/111
• feat: parallelize bootstrap generation and harden SEA startup by @mmorales-post in https://github.com/postman-cs/postman-bootstrap-action/pull/100
• test: rebind fanout evidence after squash by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/113
• fix: wait for canonical fanout links by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/115
• perf(ci): accelerate Windows parity gate by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/114
• chore: prepare v2.10.8 by @jaredboynton in https://github.com/postman-cs/postman-bootstrap-action/pull/118

Full Changelog: https://github.com/postman-cs/postman-bootstrap-action/compare/v2...v2.10.8

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the linking of services discovered by Postman Insights to a Postman workspace and Git repository, ensuring they are linked with collections, repo links, and live telemetry. It requires human-user credentials and a specific region for the insights agent discovery mode. The action does not deploy the agent or perform other setup tasks but provides a link mechanism once the service is discovered.

What’s Changed

What’s Changed

• fix: extract staged package identity cleanly by @jaredboynton in https://github.com/postman-cs/postman-insights-onboarding-action/pull/58

Full Changelog: https://github.com/postman-cs/postman-insights-onboarding-action/compare/v2.1.5...v2.1.6

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of exporting Postman collections and environments into a repository, setting up CI, mocking servers, and monitors around them. It solves the problem of managing API assets (collections, environments, mocks, monitors) in a centralized location within a Git repository. Key capabilities include:

• Exporting Postman assets to .postman directory
• Generating a CI workflow file based on selected assets
• Setting up mock servers for test environments
• Creating or updating monitors for asset health
• Integrating with GitHub Actions for continuous integration
• Using Postman API key or service token to authenticate operations

What’s Changed

What’s Changed

• fix: configure registry for release install by @jaredboynton in https://github.com/postman-cs/postman-repo-sync-action/pull/97

Full Changelog: https://github.com/postman-cs/postman-repo-sync-action/compare/v2.1.12...v2.1.13

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 20.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of identifying quantum-vulnerable cryptographic vulnerabilities in codebases using qScan, a CLI tool that scans various programming languages for RSA, (EC)DH, ECDSA, and EdDSA implementations. The action uploads SARIF formatted reports to GitHub Checks and fails the build if any new quantum-vulnerable crypto is found, providing an integrated solution for post-quantum readiness testing in CI environments.

What’s Changed

Full Changelog: https://github.com/quantakrypto/pqc-tools/compare/v0.5.0...v1

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the creation of Unity .unitypackage files by downloading a precompiled native CLI, verifying its integrity, and executing it to package assets. It simplifies the process of bundling Unity projects into portable formats across different platforms. The action supports Linux, macOS, and Windows, with options for customizing the project directory, asset files, and output path.

What’s Changed

1.0.0 (2026-07-23)

🚀 Features

• Generate deterministic .unitypackage archives on Linux, macOS, and Windows.
• Download the native unitypackage CLI and verify it using SHA-256 checksums.
• Support current inputs and compatibility aliases for existing workflows.
• Run without shell interpolation or npm runtime dependencies.

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Redis Repo Memory is a GitHub Action that automates the process of finding semantically related pull requests, issues, and commits from repository history. By using OpenAI’s text-embedding-3-small API to embed the context of each commit or PR, it searches a Redis vector index for similar prior work and posts the results as a comment on pull requests and a status update in push events. The action helps teams stay informed about what has been done before by providing insights into related content across their repository history.

What’s Changed

Remove npm caching and bump node-version to 24

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of assessing project health by scanning for common best-practice files and calculating a health score. It identifies various checks such as existence of README, LICENSE, tests, CONTRIBUTING guide, code of conduct, GitHub Actions configuration, security policy, and Dependabot setup. The tool posts the results as a PR comment and job summary, allowing users to monitor and improve project health.

What’s Changed

First release of Project Health Score.

Checks: README, LICENSE, tests, CONTRIBUTING, Code of Conduct, GitHub Actions config, Security Policy, Dependabot enabled.

• Posts a PR comment with the score, updates it on new commits
• Writes a job summary
• Outputs: score, passed-count, total-count

• This action is used across all versions by 7 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Summary:

rumdl is a high-performance Markdown linter and formatter written in Rust that provides over 76 lint rules to ensure consistency and best practices in Markdown files. It offers built-in formatting capabilities, automatic fixes, multi-flavor support, and zero dependencies. With its focus on speed and user experience, rumdl is suitable for both developers and CI/CD pipelines, making it a versatile tool for Markdown quality assurance.

What’s Changed

Added

• flavor: add Hugo flavor and skip block attribute lists in blanks-around rules (e6bb033)

Fixed

• md044: recognize indented HTML comments so links and code escape the rule (3d6191c)
• md013: keep an attr list whole inside a wrapped span (e06f03d)
• md013: keep a reference-style link whole inside a wrapped span (de42709)
• md013: wrap an over-long span whose whole content is another span (f6c7c9c)

Downloads

Installation

Using uv (Recommended)

uv tool install rumdl

Using pip

pip install rumdl

Using pipx

pipx install rumdl

Direct Download

Download the appropriate binary for your platform from the table above, extract it, and add it to your PATH.

• This action is used across all versions by ? repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

AgentAuditKit is a security scanner specifically designed to audit AI agent pipelines. It automates the detection of misconfigurations, hardcoded secrets, tool poisoning, rug pulls, trust boundary violations, and tainted data flows across various 13 agent platforms. Unlike hosted scanners that rely on LLMs for judgment, AgentAuditKit runs fully offline and deterministically, ensuring that findings are consistent across re-runs and audits. It also produces auditor-ready compliance-evidence packs in SARIF format along with PDF reports mapped to 13 security frameworks.

What’s Changed

Installation

pip:

pip install agent-audit-kit==v0.3.58

Docker:

docker pull ghcr.io/sattyamjjain/agent-audit-kit:v0.3.58

GitHub Action:

- uses: sattyamjjain/agent-audit-kit@v0.3.58
  with:
    fail-on: high

Supply chain

• rules.json — deterministic rule bundle
• rules.json.sha256 — trusted digest
• sbom.cdx.json / sbom.spdx.json — CycloneDX + SPDX SBOM
• *.sigstore — Sigstore keyless signatures (verify with agent-audit-kit verify-bundle)

Full Changelog: https://github.com/sattyamjjain/agent-audit-kit/compare/v0.3.57...v0.3.58

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Argus is a GitHub Action that automates code review by running multiple specialized AI reviewers in parallel and using an evidence-based curator to verify findings before posting review comments on pull requests. It optimizes for recall instead of precision, ensuring more real bugs are caught while minimizing false positives. The action supports various LLMs and provides markdown-based custom lenses for tailored reviews.

What’s Changed

What’s Changed

• fix(config): replace deprecated Gemini model names by @sibinms in https://github.com/sibinms/argus/pull/37
• feat(posting): reply-aware re-curation, drop the rolling summary by @sibinms in https://github.com/sibinms/argus/pull/36
• Release v1.2.25 by @sibinms in https://github.com/sibinms/argus/pull/38

Full Changelog: https://github.com/sibinms/argus/compare/v1.2.24...v1.2.25

• This publisher is shown as ‘verified’ by GitHub.

• This action is used across all versions by 8 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Socket Basics automates comprehensive security scanning using SAST, secrets detection, container scanning, and more. It normalizes outputs into Socket’s standardized format and delivers consolidated results through notification channels. Users can configure policies in the Socket Dashboard, eliminating the need for GitHub Actions workflow changes. The action supports zero configuration, unified scanning across various technologies, PR comments, centralized management, and a comprehensive guide on installation methods.

What’s Changed

What’s Changed

🚀 New Features

• feat: add diff-only scan scoping by @lelia in https://github.com/SocketDev/socket-basics/pull/77

📦 Dependencies

• chore(deps): bundle Dependabot updates + harden dependency review workflows by @lelia in https://github.com/SocketDev/socket-basics/pull/78

📚 Documentation

• chore(license): add nonempty license (LICENSE.md) by @ammkrn in https://github.com/SocketDev/socket-basics/pull/79

🔧 Other Changes

• fix: warn users when socket_org is missing by @dc-larsen in https://github.com/SocketDev/socket-basics/pull/23
• feat: add local ignore overrides for rule IDs + filepaths by @lelia in https://github.com/SocketDev/socket-basics/pull/59
• fix: honor action ‘ignore’ when generating notifications by @dc-larsen in https://github.com/SocketDev/socket-basics/pull/83
• fix: improve custom SAST rule activation, filtering semantics + config observability by @lelia in https://github.com/SocketDev/socket-basics/pull/61
• docs: document required API token scopes for Socket Basics by @dc-larsen in https://github.com/SocketDev/socket-basics/pull/68
• feat: add log statement indicating config source by @dc-larsen in https://github.com/SocketDev/socket-basics/pull/18
• fix: restore standard LICENSE filename (revert .md rename from #79) by @lelia in https://github.com/SocketDev/socket-basics/pull/88

New Contributors

• @ammkrn made their first contribution in https://github.com/SocketDev/socket-basics/pull/79

Full Changelog: https://github.com/SocketDev/socket-basics/compare/v2.0.3...v2.1.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action is a RustScript interpreter that automates the process of running and validating Rust scripts without the need to compile them fully. It supports executing and checking Rust code directly from files or snippets, providing a practical way to prototype and test small scripts without waiting for full compilation times.

What’s Changed

Full Changelog: https://github.com/VladasZ/rustscript/compare/v0.2...v0.2.1

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

AI-Driven ADR Enforcer automates the auditing of incoming Pull Requests against Architecture Decision Records (ADRs) by leveraging LLMs to ensure architectural compliance. It dynamically reads local ADR files, evaluates code diffs in real-time, and provides inline suggestions for fixing architectural violations, thus preventing drift and technical debt accumulation. The Action offers self-healing features, automated quality gates, and a safety risk assessment pipeline for small changes.

What’s Changed

Full Changelog: https://github.com/y-matsuo081991/ai-adr-enforcer/compare/v1.1.6...v1.1.7

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action is a static analysis tool that identifies discrepancies between environment variables declared in .env files and their usage in various programming languages, Dockerfiles, and Docker Compose files. It helps detect potential security risks by identifying hardcoded secrets and cross-checks variables across different sources to ensure consistency. The action supports parallel scanning and incremental caching for efficient performance.

What’s Changed

capture 2.0.0

A static analysis CLI tool that identifies mismatches between environment variables declared in .env files, Dockerfiles, and source code.

Installation

Download the appropriate binary for your platform below, or install via Go:

go install github.com/yhaliwaizman/capture/cmd/capture@v2.0.0

Changelog

• 5829a8b895e0b17f693cb8948233dd5840bfda71 feat(ci): add official GitHub Action for capture scan
• 66d296f23d4884c3dc8f74d74fb5a9312f89e658 chore(release): first major good release
• d9aaf67a253c86b38f24a6e8c73f4c6d0c06a2c1 chore(main): release 2.0.0
• 703cd3644be4b421faa0ac483f170166c06a9719 Merge pull request #40 from YhaliWaizman/release-please–branches–main

What’s Changed

See the CHANGELOG for details.

Full Changelog: https://github.com/yhaliwaizman/capture/compare/v1.7.0...v2.0.0

• This action is used across all versions by 1 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action runs the Pi coding agent as a small, composable tool, executing one non-interactive prompt and streaming the response to the job log while exposing it to later steps. It allows users to specify a prompt, API key, provider, model, and optional parameters like thinking level and tools.

What’s Changed

2.0.1 (2026-07-23)

Bug Fixes

• harden Pi setup and execution phases (92d7c14)

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Agent Lint is a Rust-based linter designed to validate various configuration files used by Claude Code, Cursor, Codex, and standalone MCP. It provides lint rules, two modes (Basic and Plugin), configurable suppression or downgrade of rules, focused execution, GitHub Action integration, cross-platform binaries, and CLI options for linting and diagnosing issues in agent configurations.

What’s Changed

What’s Changed

• Introduce a renderer-independent structured diagnostic model by @zhupanov in https://github.com/zhupanov/agent-lint/pull/206
• Restore documentation and consistency gates by @zhupanov in https://github.com/zhupanov/agent-lint/pull/207
• Add focused rule execution with –only by @zhupanov in https://github.com/zhupanov/agent-lint/pull/208
• Add versioned JSON diagnostic output by @zhupanov in https://github.com/zhupanov/agent-lint/pull/209
• Fix bare extension path classification by @zhupanov in https://github.com/zhupanov/agent-lint/pull/210
• Normalize live instruction prose across AGENTS.md and Cursor rules by @zhupanov in https://github.com/zhupanov/agent-lint/pull/211
• Refine Q002 safety prohibition handling by @zhupanov in https://github.com/zhupanov/agent-lint/pull/212
• Add end-to-end CLI conformance corpus by @zhupanov in https://github.com/zhupanov/agent-lint/pull/213
• Add agent stop-control warning by @zhupanov in https://github.com/zhupanov/agent-lint/pull/214
• Add Q005 unbounded retry prompt rule by @zhupanov in https://github.com/zhupanov/agent-lint/pull/215
• Add Q006 prompt-output-conflict rule (issue 204) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/217
• Add overlap warnings for agent and skill routing descriptions by @zhupanov in https://github.com/zhupanov/agent-lint/pull/218
• Add Q005 CLI conformance coverage by @zhupanov in https://github.com/zhupanov/agent-lint/pull/224
• Fix overlap checks for invalid YAML descriptions by @zhupanov in https://github.com/zhupanov/agent-lint/pull/225
• Fix A029 example and descriptive controls by @zhupanov in https://github.com/zhupanov/agent-lint/pull/226
• Fix Q006 output-shape classification and conflict pairs by @zhupanov in https://github.com/zhupanov/agent-lint/pull/227
• Fix bare extension classification without a fixed allowlist by @zhupanov in https://github.com/zhupanov/agent-lint/pull/231
• Fix A029 parsed agent tool declarations by @zhupanov in https://github.com/zhupanov/agent-lint/pull/232
• Fix Q005 maxTurns YAML handoff by @zhupanov in https://github.com/zhupanov/agent-lint/pull/233
• Fix Q006 clause classification false positives by @zhupanov in https://github.com/zhupanov/agent-lint/pull/267
• Fix audit JSON and description overlap gaps by @zhupanov in https://github.com/zhupanov/agent-lint/pull/268
• Harden shared inline-path classification and probes by @zhupanov in https://github.com/zhupanov/agent-lint/pull/270
• fix: unify retry and stop-control recognition by @zhupanov in https://github.com/zhupanov/agent-lint/pull/269
• Fix prompt analysis on supported instruction surfaces by @zhupanov in https://github.com/zhupanov/agent-lint/pull/274
• Retire unsupported S012/S013 skill-name rules by @zhupanov in https://github.com/zhupanov/agent-lint/pull/281
• Fix canonical skill-name validation across surfaces by @zhupanov in https://github.com/zhupanov/agent-lint/pull/282
• Fix Claude MCP remote transport validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/298
• Align plugin manifest contract with Claude Code by @zhupanov in https://github.com/zhupanov/agent-lint/pull/299
• Narrow S033 vague names and retire S049 gerund enforcement by @zhupanov in https://github.com/zhupanov/agent-lint/pull/301
• Extend M012/M013 plugin component path coverage by @zhupanov in https://github.com/zhupanov/agent-lint/pull/313
• Fix manifest diagnostic precision by @zhupanov in https://github.com/zhupanov/agent-lint/pull/316
• Deep-validate marketplace plugin entries (M009/M019/M021) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/317
• Make MCP discovery and P-rule dispatch platform-aware by @zhupanov in https://github.com/zhupanov/agent-lint/pull/323
• Fix MCP structural validation and rule independence by @zhupanov in https://github.com/zhupanov/agent-lint/pull/331
• Add structured metadata to MCP diagnostics by @zhupanov in https://github.com/zhupanov/agent-lint/pull/332
• Make P018/P019 secret and command analysis syntax-aware by @zhupanov in https://github.com/zhupanov/agent-lint/pull/338
• Fix hook command path extraction and Basic autofix by @zhupanov in https://github.com/zhupanov/agent-lint/pull/347
• Fix plugin hook configuration discovery by @zhupanov in https://github.com/zhupanov/agent-lint/pull/351
• Align userConfig validation with current Claude Code schema by @zhupanov in https://github.com/zhupanov/agent-lint/pull/349
• Split and harden E001 without exposing email values by @zhupanov in https://github.com/zhupanov/agent-lint/pull/363
• Remove repository-specific K001 Slack fallback policy by @zhupanov in https://github.com/zhupanov/agent-lint/pull/357
• Fix script reference integrity rules by @zhupanov in https://github.com/zhupanov/agent-lint/pull/365
• Fix S030/S036/S072/S073 skill structural precision by @zhupanov in https://github.com/zhupanov/agent-lint/pull/370
• Fix Q004 live prose overlap detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/371
• Extend S031/S032 to .agents/skills and .cursor/skills by @zhupanov in https://github.com/zhupanov/agent-lint/pull/372
• Fix canonical agent YAML field validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/373
• Broaden body recognizers for S041/S046/S047/S055 by @zhupanov in https://github.com/zhupanov/agent-lint/pull/380
• Fix Codex config schema validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/383
• Narrow I004 generic-only guidance and remove unsound I005 by @zhupanov in https://github.com/zhupanov/agent-lint/pull/379
• Fix canonical multiline skill descriptions by @zhupanov in https://github.com/zhupanov/agent-lint/pull/389
• Broaden G005 to GitHub-supported SECURITY.md locations by @zhupanov in https://github.com/zhupanov/agent-lint/pull/375
• Fix path-root classification before PWD autofix by @zhupanov in https://github.com/zhupanov/agent-lint/pull/364
• S055: recursively inspect shebang scripts and assign diagnostics to the offending file by @zhupanov in https://github.com/zhupanov/agent-lint/pull/395
• Align S015 with Claude Code listing cap by @zhupanov in https://github.com/zhupanov/agent-lint/pull/394
• I002: make AGENTS.md credential detection complete, source-aware, and non-leaking by @zhupanov in https://github.com/zhupanov/agent-lint/pull/393
• Model Q006 response scopes as an inherited heading tree by @zhupanov in https://github.com/zhupanov/agent-lint/pull/386
• Fix agent frontmatter field validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/406
• Recognize A029/Q005 operative gates through Markdown emphasis by @zhupanov in https://github.com/zhupanov/agent-lint/pull/407
• Escape control characters in human-readable diagnostic output by @zhupanov in https://github.com/zhupanov/agent-lint/pull/405
• Fix typed Codex MCP secret detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/408
• Fix S059 script flag validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/409
• Fix S016/S018/S054 description heuristic false positives by @zhupanov in https://github.com/zhupanov/agent-lint/pull/414
• Fix canonical frontmatter autofix scope by @zhupanov in https://github.com/zhupanov/agent-lint/pull/416
• Fix S022 backslash path detection and autofix by @zhupanov in https://github.com/zhupanov/agent-lint/pull/417
• Broaden H023 dangerous-command pattern coverage by @zhupanov in https://github.com/zhupanov/agent-lint/pull/418
• L006: validate actionable npm commands in shell fences by @zhupanov in https://github.com/zhupanov/agent-lint/pull/419
• S031: stop flagging/rewriting XML-namespace identifiers and reserved-name hosts; single shared checker/autofix classifier by @zhupanov in https://github.com/zhupanov/agent-lint/pull/420
• Fix nested Cursor rule discovery by @zhupanov in https://github.com/zhupanov/agent-lint/pull/421
• Fix markdown fence and XML structure parsing by @zhupanov in https://github.com/zhupanov/agent-lint/pull/426
• Improve S/X diagnostic precision for shared refs and frontmatter by @zhupanov in https://github.com/zhupanov/agent-lint/pull/427
• Narrow D003/G006/G007 to syntactic unfinished-work markers by @zhupanov in https://github.com/zhupanov/agent-lint/pull/429
• fix: align Cursor hooks validation contract by @zhupanov in https://github.com/zhupanov/agent-lint/pull/428
• T001/T002: validate rendered PR URLs and reject ignored repository channelsEnabled by @zhupanov in https://github.com/zhupanov/agent-lint/pull/430
• Fix A012/A013 canonical agent contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/432
• S044: gate context words on word boundaries, not substrings by @zhupanov in https://github.com/zhupanov/agent-lint/pull/431
• Share operative/example scopes across Q001-Q003 and scope Q002 alternatives (#359) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/433
• Fix M021 to match kebab-case against raw marketplace names by @zhupanov in https://github.com/zhupanov/agent-lint/pull/439
• Fix S032 source-positioned secret detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/438
• A005-A007: self-activate the larch template convention and parse Markdown semantics by @zhupanov in https://github.com/zhupanov/agent-lint/pull/442
• D004/L001-L004: build a complete, safe, source-relative instruction import graph by @zhupanov in https://github.com/zhupanov/agent-lint/pull/441
• Unify shared model/tool vocabularies (S063/A014, S040/A019/A020) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/440
• U009: flag userConfig defaults that ship secrets (#412) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/444
• S021 autofix: follow the shared consecutive-bash policy (#326) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/445
• Fix S037 file reference recognition by @zhupanov in https://github.com/zhupanov/agent-lint/pull/446
• CX013: narrow Slack signature class to reject backslashes (#425) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/448
• Fix Q006 hard-wrapped conditional routing (#399) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/447
• Fix script reference extraction gaps by @zhupanov in https://github.com/zhupanov/agent-lint/pull/451
• Sync MCP P026/P027/P009/P017 with current Claude/Cursor contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/449
• Validate marketplace component paths by @zhupanov in https://github.com/zhupanov/agent-lint/pull/452
• G008: make inline-body checks GitHub-command-aware and shell-aware by @zhupanov in https://github.com/zhupanov/agent-lint/pull/454
• Fix CU016 Cursor environment schema validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/456
• Fix masked prose validation rules by @zhupanov in https://github.com/zhupanov/agent-lint/pull/455
• Report malformed hook configurations by @zhupanov in https://github.com/zhupanov/agent-lint/pull/458
• Test Cursor frontmatter prompt recovery by @zhupanov in https://github.com/zhupanov/agent-lint/pull/457
• CU014: optional Cursor subagent fields and A030 Cursor namespace by @zhupanov in https://github.com/zhupanov/agent-lint/pull/459
• I003/D005/L005/S062: structured Markdown refs and safe path resolution by @zhupanov in https://github.com/zhupanov/agent-lint/pull/460
• Fix Cursor skill field compatibility and discovery by @zhupanov in https://github.com/zhupanov/agent-lint/pull/462
• Agent discovery: scan .claude/agents and plugin agents/ recursively (#321) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/463
• CX046-CX063: rebuild Codex plugin manifest linting by @zhupanov in https://github.com/zhupanov/agent-lint/pull/461
• G009-G011: replace line regexes with sound shell and awk analysis by @zhupanov in https://github.com/zhupanov/agent-lint/pull/464
• S014/S034: enforce the Agent Skills description contract by @zhupanov in https://github.com/zhupanov/agent-lint/pull/465
• Fix Codex diagnostic metadata contract by @zhupanov in https://github.com/zhupanov/agent-lint/pull/467
• Make D001/D002 Always-mode with precise Markdown docs refs by @zhupanov in https://github.com/zhupanov/agent-lint/pull/468
• Honor plugin skill export surfaces by @zhupanov in https://github.com/zhupanov/agent-lint/pull/466
• CX060: parse YAML semantically across every Codex skill surface by @zhupanov in https://github.com/zhupanov/agent-lint/pull/469
• Fix Codex project-document discovery and budget rules by @zhupanov in https://github.com/zhupanov/agent-lint/pull/470
• O001-O006: align output-style linting with Claude runtime by @zhupanov in https://github.com/zhupanov/agent-lint/pull/471
• Fix Q005 retry bound association by @zhupanov in https://github.com/zhupanov/agent-lint/pull/473
• Fix S058 Skill invocation detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/474
• Fix S060/S061 fence shell-command precision by @zhupanov in https://github.com/zhupanov/agent-lint/pull/475
• A002/A003/A008-A011: use canonical agent YAML and re-severitize description heuristics by @zhupanov in https://github.com/zhupanov/agent-lint/pull/477
• Fix H023 Git option and shell wrapper detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/482
• M014-M017: share plugin-field validation across manifests and enforce current LSP/channel shapes by @zhupanov in https://github.com/zhupanov/agent-lint/pull/478
• Read skill field-type frontmatter through canonical YAML (#341) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/480
• Extend P019 threat matrix for PowerShell prefixes, root globs, argv joining, and headersHelper by @zhupanov in https://github.com/zhupanov/agent-lint/pull/484
• CU002-CU008: model Cursor MDC frontmatter as four activation states by @zhupanov in https://github.com/zhupanov/agent-lint/pull/476
• Add M024 marketplace whitespace rule by @zhupanov in https://github.com/zhupanov/agent-lint/pull/483
• S028/S069: align argument-hint cross-checks with merged command-argument contract by @zhupanov in https://github.com/zhupanov/agent-lint/pull/472
• R001-R003: recursively validate Claude rules with runtime-compatible paths semantics by @zhupanov in https://github.com/zhupanov/agent-lint/pull/481
• S033: complete domainless denylist and attach name-field metadata by @zhupanov in https://github.com/zhupanov/agent-lint/pull/485
• A031 agent-name-duplicate: flag duplicate agent names in .claude/agents (identity collision) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/486
• Fix S065 agent reference resolution by @zhupanov in https://github.com/zhupanov/agent-lint/pull/489
• G005: accept security-md-missing as a legacy alias (#411) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/487
• Fix S068 inline injection overflow detection by @zhupanov in https://github.com/zhupanov/agent-lint/pull/488
• Fix recursive S030 and S048 skill coverage by @zhupanov in https://github.com/zhupanov/agent-lint/pull/490
• S040/S045/S067: honor documented allowed-tools/disallowed-tools forms; retire S045 by @zhupanov in https://github.com/zhupanov/agent-lint/pull/491
• test: pin retired rule identifiers by @zhupanov in https://github.com/zhupanov/agent-lint/pull/492
• CX026/CX030: validate nested approvals_reviewer and default_tools_approval_mode sites by @zhupanov in https://github.com/zhupanov/agent-lint/pull/493
• docs: record Cursor contract provenance by @zhupanov in https://github.com/zhupanov/agent-lint/pull/498
• Validate Codex profile configuration values by @zhupanov in https://github.com/zhupanov/agent-lint/pull/494
• O001-O006: lint plugin-shipped output styles (root output-styles/ and manifest outputStyles paths) in Plugin mode by @zhupanov in https://github.com/zhupanov/agent-lint/pull/495
• M010/M011: require usable enrichment values without parent cascades by @zhupanov in https://github.com/zhupanov/agent-lint/pull/499
• CHANGELOG: record landed manifest-rule campaign (M001-M021) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/501
• Fix hook script invocation classification by @zhupanov in https://github.com/zhupanov/agent-lint/pull/497
• Fix manifest root and plugin name validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/500
• Fix P018 MCP credential reference coverage by @zhupanov in https://github.com/zhupanov/agent-lint/pull/505
• S041 (fork-no-task): read context through canonical YAML by @zhupanov in https://github.com/zhupanov/agent-lint/pull/504
• Reverify Cursor contracts against cursor.com/docs by @zhupanov in https://github.com/zhupanov/agent-lint/pull/507
• Fix #437: refresh Claude Code vocabularies by @zhupanov in https://github.com/zhupanov/agent-lint/pull/503
• Fix focused JSON output for retired rule selectors by @zhupanov in https://github.com/zhupanov/agent-lint/pull/564
• fix: close S032 command substitution bypass by @zhupanov in https://github.com/zhupanov/agent-lint/pull/565
• Fix S058 canonical YAML tool gate by @zhupanov in https://github.com/zhupanov/agent-lint/pull/566
• fix: balance R001 brace expansion so nested empties fail closed by @zhupanov in https://github.com/zhupanov/agent-lint/pull/567
• fix: stop Q rules from scanning unterminated and BOM frontmatter by @zhupanov in https://github.com/zhupanov/agent-lint/pull/568
• fix: cover hook command runtime boundaries by @zhupanov in https://github.com/zhupanov/agent-lint/pull/571
• Fix plugin and marketplace validation drift by @zhupanov in https://github.com/zhupanov/agent-lint/pull/569
• Fix P019 command argv boundaries by @zhupanov in https://github.com/zhupanov/agent-lint/pull/570
• fix: unify runtime skill discovery consumers by @zhupanov in https://github.com/zhupanov/agent-lint/pull/572
• Fix CommonMark Markdown tokenization and link decoding by @zhupanov in https://github.com/zhupanov/agent-lint/pull/575
• fix: assign npm and grep positional roles option-awarely by @zhupanov in https://github.com/zhupanov/agent-lint/pull/573
• fix: align instruction reference discovery for imports and package scopes by @zhupanov in https://github.com/zhupanov/agent-lint/pull/574
• fix: accept documented empty and templated Claude MCP URLs (P010/P017) by @zhupanov in https://github.com/zhupanov/agent-lint/pull/576
• Fix G008-G011 shell analysis dispatch and dynamic-flow gaps by @zhupanov in https://github.com/zhupanov/agent-lint/pull/577
• fix: tighten instruction classifier grammars by @zhupanov in https://github.com/zhupanov/agent-lint/pull/578
• fix: complete Codex plugin manifest validation contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/579
• fix: keep Codex diagnostics secret-safe and budget-visible by @zhupanov in https://github.com/zhupanov/agent-lint/pull/581
• Fix JSON diagnostics losing duplicate, field, and array source locations by @zhupanov in https://github.com/zhupanov/agent-lint/pull/580
• fix: correct agent root, marker, name, and metadata classification by @zhupanov in https://github.com/zhupanov/agent-lint/pull/582
• Enforce parser-role and invalid-subtree contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/584
• Append bug-mining architectural entries by @zhupanov in https://github.com/zhupanov/agent-lint/pull/588
• Unify executable and retired identifier contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/590
• Add conformance coverage and source-policy guards by @zhupanov in https://github.com/zhupanov/agent-lint/pull/597
• test: enforce diagnostic safety corpus by @zhupanov in https://github.com/zhupanov/agent-lint/pull/599
• Add AST-backed validator purity and traversal lints by @zhupanov in https://github.com/zhupanov/agent-lint/pull/606
• Fix live rule registry accounting by @zhupanov in https://github.com/zhupanov/agent-lint/pull/607
• test: inventory validator peer dependencies by @zhupanov in https://github.com/zhupanov/agent-lint/pull/608
• Fix S059 modular Python CLI flag validation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/611
• fix: S030 ignore cache artifacts and harness-owned fixtures by @zhupanov in https://github.com/zhupanov/agent-lint/pull/609
• fix: stop S032 flagging ordinary token identifiers by @zhupanov in https://github.com/zhupanov/agent-lint/pull/610
• Fix Markdown control scope classifiers by @zhupanov in https://github.com/zhupanov/agent-lint/pull/613
• Enforce CWD test guards and serialization by @zhupanov in https://github.com/zhupanov/agent-lint/pull/612
• docs: record architectural contracts by @zhupanov in https://github.com/zhupanov/agent-lint/pull/614
• Generalize contract matrix admission by @zhupanov in https://github.com/zhupanov/agent-lint/pull/617
• fix: keep UTF-8 scalars intact in clause splitting and sourcepos ranges by @zhupanov in https://github.com/zhupanov/agent-lint/pull/618
• fix: unify script reference classification across G002-G004 by @zhupanov in https://github.com/zhupanov/agent-lint/pull/619
• test: add text/JSON renderer parity invariant by @zhupanov in https://github.com/zhupanov/agent-lint/pull/620
• test: cover UTF-8 clause scanning regression by @zhupanov in https://github.com/zhupanov/agent-lint/pull/622
• Fix Comrak column semantics for Unicode prose by @zhupanov in https://github.com/zhupanov/agent-lint/pull/624
• test: enforce architecture and rule/autofix consistency by @zhupanov in https://github.com/zhupanov/agent-lint/pull/623
• Fix S059 literal Python registry dispatch by @zhupanov in https://github.com/zhupanov/agent-lint/pull/629
• fix: recognize S030 references ending in sentence punctuation by @zhupanov in https://github.com/zhupanov/agent-lint/pull/630
• fix: recognize Python-owned skill fixtures by @zhupanov in https://github.com/zhupanov/agent-lint/pull/632
• Fix G004 Python-composed script reachability by @zhupanov in https://github.com/zhupanov/agent-lint/pull/631
• fix: accept End and Schedule Q002 alternatives by @zhupanov in https://github.com/zhupanov/agent-lint/pull/637
• Fix G004 Python subprocess reachability by @zhupanov in https://github.com/zhupanov/agent-lint/pull/638
• Fix S030 Python fixture string ownership by @zhupanov in https://github.com/zhupanov/agent-lint/pull/645
• Fix S059 imported Python literal registries by @zhupanov in https://github.com/zhupanov/agent-lint/pull/647
• Fix Python hygiene command edges by @zhupanov in https://github.com/zhupanov/agent-lint/pull/649
• Fix G003 command-substitution interpreter classification by @zhupanov in https://github.com/zhupanov/agent-lint/pull/651

Full Changelog: https://github.com/zhupanov/agent-lint/compare/v3...v4.0.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Syscribe is a tool that converts SysMLv2 models into human-readable Markdown files with YAML frontmatter, making them version-controlled and traceable. It automates the process of generating requirements, test cases, architecture decision records, and safety analysis reports from natural language descriptions using large language models (LLMs). The action supports various element types and provides features for cross-repository composition, security analysis, and IEC 62443 zones & conduits.

What’s Changed

Full Changelog: https://github.com/sjames/syscribe/compare/v0...v0.32.0

• This action is used across all versions by 1 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Muninn is an open-source security scanning tool designed to integrate with GitHub Actions workflows. It automates the process of identifying vulnerabilities across various tools, normalizes their output, and presents findings in a unified format as comments on pull requests, SARIF uploads, and structured JSON. Muninn uses eight popular vulnerability scanners to detect secrets, pipeline misconfigurations, syntax errors, supply chain risks, application SAST issues, dependency vulnerabilities, container image vulnerabilities, and infrastructure-as-code misconfigurations, providing a comprehensive security scan of GitHub Actions pipelines and self-hosted CI environments.

What’s Changed

Changelog

[0.3.6] - 2026-07-22

Changed

• Docker image scanner updates: zizmor 1.28.0 (replaces yanked 1.27.0); GitPython floor raised to >=3.1.52 for checkov-transitive GHSA highs (checkov remains at 3.2.531 pending aiohttp cap lift).
• Go toolchain bumped to 1.26.5.

[0.3.5] - 2026-07-20

Changed

• Docker image scanner updates: semgrep 1.170.0, zizmor 1.27.0 (checkov remains at 3.2.531 pending aiohttp cap lift).

[0.3.4] - 2026-07-04

Changed

• Docker image scanner updates: osv-scanner 2.4.0, trivy 0.72.0, semgrep 1.168.0, zizmor 1.26.1 (checkov remains at 3.2.531 pending aiohttp cap lift).

[0.3.3] - 2026-06-17

Changed

• Trivy default severity is now all levels (UNKNOWN through CRITICAL) instead of CRITICAL and HIGH only. osv-scanner and trivy now overlap on medium/low advisories by default so cross-scanner dedup and Detected by work without extra config. Consumers can narrow the Trivy scan with scanners.trivy.severity; fail-on still controls which findings fail the run.

[0.3.2] - 2026-06-16

Fixed

• Suppressions with tool and/or rule-id are now applied. Previously only id (path substring) and fingerprint matchers worked; tool+rule-id entries parsed from muninn.yml but silently no-op’d.

[0.3.1] - 2026-06-16

Fixed

• Poutine v1.x JSON parsing: findings from poutine 1.1.6+ (rule_id, meta, rules, blobshas) now populate title, rule, and file in PR comments instead of empty shells (File: :0, `Rule: ``) (#41).
• Actionlint PR comments: fall back to kind (e.g. expression) when rule.name is absent; omit empty Rule lines.
• Poutine injection findings: render injection_sources as formatted Sources instead of plain meta.details text.

Changed

• PR comment layout: shared field helpers; non-dependency findings follow File → Rule → optional extras → description; single-scanner dependency findings use File instead of a redundant Source line.

[0.3.0] - 2026-06-16

Added

• Cross-scanner deduplication by advisory id: findings that report the same CVE/GHSA for the same package from different scanners (e.g. OSV-Scanner from a lockfile and Trivy from a container layer) are now collapsed into a single finding. The contributing scanners are recorded in a new detected_by field (surfaced in the JSON report, the PR comment’s “Detected by” line, and a detectedBy SARIF result property). A CVE is preferred over GHSA so the same vulnerability converges on one id across scanners (#27).
• Richer dependency finding rendering: aggregated dependency findings now appear under a neutral [dependency] heading (instead of a single scanner’s name) with Package, Advisory (including the shared CVE), Detected by, and a Sources list showing where each scanner observed it. A new sources field on the finding (per-scanner tool + file) backs the JSON report (#27).

Fixed

• PR comment rendering: scanner descriptions are flattened to a single line and their Markdown (code fences, headings) neutralized, so an unbalanced ``` fence can no longer swallow later findings and the footer into a code block.

[0.2.0] - 2026-06-15

Supply-chain hardening for the scanner image and signed, verifiable releases (closes #30).

Added

• Pinned every bundled binary scanner to an exact version with SHA256 checksum verification in the Docker image — gitleaks, zizmor, actionlint, poutine, osv-scanner, trivy (#31)
• Hash-locked the pip-installed scanners (semgrep, checkov, zizmor) via a fully pinned, multi-arch requirements-scanners.txt installed with pip --require-hashes (#33)
• Renovate configuration to auto-PR scanner version bumps, with a CI job that refreshes the pinned checksums (#32)
• Keyless (OIDC) cosign signing of the published container image and of the release binary checksums (Sigstore bundle checksums.txt.sigstore.json) (#34)
• SBOM (SPDX) attached to every release and as an image attestation (#34)
• Max-mode SLSA build provenance attestation on the container image (#34)
• “Verifying releases” instructions in the README (#34)

Changed

• Pinned checkov to 3.2.531 (from 3.3.1) so its dependency tree resolves the patched aiohttp 3.14.1 and drops the unfixable python-ecdsa Minerva dependency that checkov 3.3.x introduced. Revisit when a newer checkov lifts its aiohttp<3.14 cap (#33)

[0.1.0] - 2026-06-14

Added

• 8 security scanners: gitleaks, zizmor, actionlint, poutine, semgrep, osv-scanner, trivy, checkov
• Unified Finding schema with fingerprinting
• Three output formats: SARIF 2.1.0, JSON, GitHub PR comment
• GitHub Action with outputs
• Config-driven scanner behavior via muninn.yml
• Suppression management with expiry dates
• 90%+ test coverage enforced in CI
• Integration tests with real scanner binaries
• Self-scan: Muninn scans itself on every PR

Built by Skald Lab — skaldlab.dev

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

SSG is a fast static site generator written in Go that converts Markdown with YAML frontmatter into a complete website, supporting various themes, template engines, SEO features, and deployment options. It automates the process of generating clean URLs, feeds, search indexes, and more, making it suitable for building blogs, documentation sites, and landing pages efficiently.

What’s Changed

Installation

Quick Install (Linux/macOS)

curl -sSL https://raw.githubusercontent.com/spagu/ssg/main/install.sh | bash

Package Managers

• Homebrew: brew install spagu/tap/ssg
• Snap: snap install ssg
• Debian/Ubuntu: Download .deb file below
• Fedora/RHEL: Download .rpm file below

Checksums

See checksums.sha256 for file verification.

📖 Full documentation: https://github.com/spagu/ssg#readme

What’s Changed

• 1.8.12 — Mermaid diagrams + code line numbers by @spagu in https://github.com/spagu/ssg/pull/46
• fix(docs-site): readable code blocks — github-dark highlight style by @spagu in https://github.com/spagu/ssg/pull/48
• feat: strip_md_link_text — drop .md from link text at publish (GO-075) by @spagu in https://github.com/spagu/ssg/pull/49

Full Changelog: https://github.com/spagu/ssg/compare/v1.8.11...v1.8.12

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The spek GitHub Action provides a lightweight, read-only viewer for OpenSpec content, offering structured browsing with BDD syntax highlighting, task progress tracking, and full-text search. It automates the process of displaying and managing specs, changes, and tasks in a single interface, making it easier to collaborate and understand complex projects.

What’s Changed

• The IntelliJ plugin opens again on IDEs built on platform 2026.2. Opening the Tool Window threw an IDE Internal Error and left no viewer at all, so the plugin was unusable — 2026.2 moved JCEF into a bundled plugin whose classes the plugin could no longer see (#24, reported on WebStorm 2026.2 with plugin 1.9.0). The embedded webview is restored, and JCEF being unavailable for any reason now degrades to the external-browser fallback instead of crashing — which is what makes the next platform change of this kind an inconvenience rather than an outage.
• The timeline’s “group by topic” works when worktree aggregation is on. Every change landed in the “(no topic)” lane, which read as though the repo had no spec relationships at all. Nothing looked broken — the chart rendered exactly as it does when the grouping is genuinely empty. Repos with a single worktree were unaffected (#25, reported by a downstream consumer of the @spekjs/ui package). Web and VS Code only; the IntelliJ plugin has no worktree aggregation, so it never hit this.

Also released: @spekjs/ui@1.1.0 — the same grouping fix for registry consumers, plus a new exported changeNodeSlug(node) helper for resolving graph change node ids. See packages/ui/CHANGELOG.md.

• This action is used across all versions by 4 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action initializes Nix-based repositories by automating common setup tasks such as creating a GitHub App Token, checking out the repository, setting up Git user information, configuring an optimal Nix environment, installing Nix, and setting the Nix configuration from a Flake. It also supports caching with niks3 for better performance on self-hosted runners. The action is designed to run efficiently (< 1 minute) and can be used with various runner types (GitHub Actions, Gitea, Forgejo).

What’s Changed

• feat(action): use node for nix develop env export (718723a4b53683c78e4dfcfa2f1da4a78264b171)
• feat(action): use vendored nix-develop script (d24036e34bc83eaa5f2637ba61d089ff8f4edaf1)
• feat: Update spotdemo4/better-checkout action to v0.14.0 (#168) (b28fd2c4d5f02afb7135720526d061c9794d4749)
• fix(tests): check requested env vars only (092ab684bd4227911e870c62c5127a2f19e17c96)
• fix(nix-develop): snapshot exported host vars for bash 3.2 (2740b4c10b72ee87614b33b6f9c6c43feeab6480)
• fix(nix-develop): avoid nounset from unset indirections (b6c3c02f35216c601bffc33f872dc4fd6e767164)
• bump: v1.59.0 -> v1.60.0 (2fc360eefced0d01fc56eaf691ab9393a94a1c60)
• chore: add top comment to shell setup (ec872f577e0291c8f52260f4ce4c283e187d9b3b)
• test(ci): validate devShell environment exports (1d68c4040201627cbcc638e596acf959b7d16429)
• chore(deps): update github actions to v7.0.1 (#167) (2d87c0572b541faad84eea63853d4175504d9c59)
• chore(deps): lock file maintenance nix inputs (#166) (2738beb2e2d1be14ada7c74133fcf7861d5982da)
• chore(deps): update github actions to v1.59.0 (#165) (57bf48f3b354214cd10358b4ee755e3d50f71a02)

• This action is used across all versions by 8 repositories.

Action Type

This is a Docker action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates static analysis of WDL documents using Sprocket, a tool for validating and linting WDL files. It provides subcommands for check, lint, and validate, allowing users to enforce coding standards and detect potential issues in their workflows. The action can be configured with input parameters such as lint rules, ignore patterns, and deny options. It also supports formatting checks using a sprocket.toml configuration file.

What’s Changed

Bumps sprocket to v0.28.0.

• This action is used across all versions by 2 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The xilo action is a self-hosted Nix binary cache written in Go. It offers single-tenant multi-user support, content-addressed chunk deduplication, and supports local or S3 storage backends. Key features include instant token revocation, no stalls on concurrent pushes, a cachix-style admin dashboard, and a 9 MB Docker image serving zstd pulls directly from stored frames.

What’s Changed

What’s Changed

• chore(ci): bump actions/setup-go from 6 to 7 in the actions group by @dependabot[bot] in https://github.com/stubbedev/xilo/pull/10
• chore(deps): bump the go-deps group with 5 updates by @dependabot[bot] in https://github.com/stubbedev/xilo/pull/11

Full Changelog: https://github.com/stubbedev/xilo/compare/v1...v1.0.10

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Sudden Agent automates tasks on GitHub workflows through programmable agents. It solves problems related to automating code reviews, issue triage, security audits, and version management by providing customizable prompts and session persistence across different actions in the workflow. The action supports various models and authentication methods for different agents.

What’s Changed

What’s Changed

• Support GPT-5.6 Codex models by @etienne-martin in https://github.com/sudden-network/agent/pull/117

Full Changelog: https://github.com/sudden-network/agent/compare/v1.14.0...v1.14.1

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Sigbound automates the process of running multiple AI coding agents on one repository in parallel, merging their work automatically. It handles conflicts by using a model to resolve them and ensures that only changes that build and pass your tests are landed. The action is designed to be flexible, allowing users to bring their own model for planning, agent execution, conflict resolution, and merge repair.

What’s Changed

Run AI coding agents in parallel on one git repo and auto-merge their verified work — nothing lands unless it builds and passes your tests.

Highlights in 0.3.0

• -verify-bisect — when the combined tree fails verify, land the verified green subset instead of losing everything
• -verify-cache — skip re-verifying trees already proven green (keyed by tree OID)
• Run manifest + sig replay — every landing is reproducible provenance
• -resume — pick up an interrupted run without re-running finished agents
• -publish — push the landed commit and open a PR/MR with your own tooling
• This GitHub Action — install a released sig, run it from typed inputs, gate your job on the outcome

Use it

- uses: surya-koritala/sigbound@v0.3.0
  with:
    agent: 'claude -p --permission-mode acceptEdits "$SIGBOUND_TASK"'
    goal: "Add CSV export, due dates, and a summary command"
    verify: "go build ./... && go test ./..."

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action calculates the hash of specified folders and outputs the result, which can be used to generate cache keys or as a checksum for folder content. It supports caching multiple folders and uses glob patterns for file selection.

What’s Changed

Breaking changes

• Folder hash values have changed. Regenerate any cache keys that depend on this action after upgrading.

Changed

• Upgraded the action and CI runtime to Node 24.
• Migrated source and tests to ES modules and Node’s native test runner.
• Replaced deprecated hashing and bundling dependencies.
• Refreshed the bundled action output and documentation.

Full Changelog: https://github.com/tankist/folder-hash/commits/v4.0.0

• This action is used across all versions by ? repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Tenzai Test GitHub Action automates AI-powered security testing of deployments directly from CI pipelines. It sends a commit-diff test against an existing Tenzai application and provides real-time feedback on vulnerabilities, automatically triggering a Tenzai Test check run in the tested commit. The action requires a production Tenzai service-account access key and app ID, validates configuration, and supports dry runs for testing purposes.

What’s Changed

What’s Changed

• add baseurl override to action by @Dor256 in https://github.com/TenzaiLtd/tenzai-github-action/pull/9

Full Changelog: https://github.com/TenzaiLtd/tenzai-github-action/compare/v1.0.1...v1.0.2

• This action is used across all versions by 11 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Soundcheck Security Review (v2) GitHub Action automates the execution of Soundcheck, an OWASP-focused static code analysis tool. It performs a security review on all changes in a pull request or a scheduled scan across the repository, providing a severity-ranked findings table and optional autofix capabilities that rewrite critical issues directly into the affected files.

What’s Changed

Bumps soundcheck to v1.17.0 and the claude CLI pin to 2.1.214.

The v1.17.0 changes on the soundcheck side:

• hotspot-mapping now covers rendering/output sinks and dependency/config manifests (previously blindspots)
• vulnerability-audit is exhaustive per hotspot — picks every applicable skill, sweeps sibling code

Benchmark against OWASP Juice Shop: reviews now produce ~2× more findings and lift strict recall from 33% → 67%, with zero hallucinations across residual findings.

No breaking changes to inputs, outputs, or workflow shape.

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

Translatize Sync is a GitHub Action that automates the synchronization of source keys and translations between a local repository and a Translatize project. It simplifies the process by providing commands to push new source keys, pull completed translations, and gate pull requests based on missing or changed strings. The action uses a branch-bound API token for authentication and supports configuration via translatize.config.json.

What’s Changed

First public release of Translatize Sync.

Sync your app’s translations with Translatize from GitHub workflows:

• push – upload new and changed source keys to a project branch
• pull – download completed translations into your locale files
• status – compare local files with the branch; gate PRs with --fail-on-missing / --fail-on-diff

Authenticates with a branch-bound Translatize API token (repo secret). Works with any repo layout via translatize.config.json.

Docs: https://translatize.com/docs/github-action

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

repo-trust-scan is a CLI and GitHub Action that statically scans repository-controlled execution surfaces, helping users identify and understand potential security risks before executing unfamiliar code. It checks tasks like automatic editor actions, agent hooks, MCP server configurations, devcontainer lifecycle commands, package install hooks, escaping symlinks, hidden Unicode in instructions, download-and-execute chains, and credential-transfer patterns. The action is useful for ensuring that repositories are secure and safe to work with, complementing existing security tools but not replacing them.

What’s Changed

Highlights

• Publishes repo-trust-scan as a reusable GitHub Action in the Security and Utilities categories.
• Adds a 21-second terminal demo and a 1280×640 social preview asset.
• Moves real scan output ahead of installation and adds clear comparisons with secret scanners, SAST, and runtime scanners.
• Adds a copyable GitHub Actions workflow.
• Updates pinned GitHub Actions dependencies.

Validation

• 25 unit tests pass (2 Windows symlink tests skipped as expected).
• Package build, action smoke test, compile check, and critical self-scan pass.
• CI passes on Ubuntu, Windows, and macOS with Python 3.10 and 3.14.

• This action is used across all versions by 0 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action scans Markdown and MDX files to find broken, redirected, insecure HTTP links, and excessively long URLs. It can be used as a CLI or as part of a continuous integration pipeline to ensure the integrity of documentation links.

What’s Changed

Full Changelog: https://github.com/URLdn/link-check/compare/v1.0.0...v1.0.1

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action difftrace automates change detection in monorepos by analyzing the uv.lock file to build a dependency graph, mapping changes to packages, and identifying transitively affected packages. It identifies which packages are directly or indirectly impacted by code changes, optimizing CI pipelines to only process relevant packages.

What’s Changed

What’s Changed

• fix: key lock fingerprints per-name-set so multi-version deps aren’t hidden by @vanandrew in https://github.com/vanandrew/difftrace/pull/13

Full Changelog: https://github.com/vanandrew/difftrace/compare/v1.5.0...v1.5.1

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The @vibgrate/cli GitHub Action provides a comprehensive local codebase intelligence tool for AI coding agents. It automates several key tasks including generating a deterministic code graph, calculating drift scores to measure how far behind the codebase is in terms of runtime dependencies and library versions, and providing ranked upgrade priorities for maintaining up-to-date dependencies. The action runs entirely on your machine, ensuring no network calls or data leaves your repository unless explicitly pushed.

What’s Changed

Vibgrate CLI 2026.722.2

Released 2026-07-22

This release introduces new commands for managing models in the vg CLI. Users can now access a live hosted model catalog and remove locally-installed models more easily.

What changed

New

• vg models catalog provides a key-free, cached list of live hosted models from OpenRouter, with options for JSON output and offline caching.
• vg models rm <name> allows for the removal of locally-installed models, printing a plan by default and requiring --yes for confirmation.

Benchmarks

Two-arm benchmark of this release against 2026.722.1, interleaved on one runner against the pinned corpus (176 metrics compared).

No regressions against the previous release.

Full report and methodology: https://vibgrate.com/cli/benchmarks

Install or update

npm install -g @vibgrate/cli
vg

Full changelog: https://vibgrate.com/changelog/cli/2026.722.2

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The Repo Settings as Code GitHub Action automates the application of declarative repository settings from a .github/settings.yml file to GitHub repositories. It provides a loud, stateless replacement for the Probot Settings app, which also handles rulesets (branch, tag, and push). The action ensures that every apply is visible and fails with API error messages, ensuring no silent failures occur. It supports setting up fine-grained Personal Access Tokens (PATs) as repository secrets, using them to manage permissions and settings declaratively in YAML format.

What’s Changed

Repo Settings as Code: First release.

Apply declarative repository settings from .github/settings.yml: a stateless replacement for the Probot Settings app.

• Sections: repository fields, labels, rulesets, branch protection, environments, autolinks, Actions permissions, workflow state, Pages, code scanning default setup, collaborators, teams, and milestones.
• Every apply is a visible workflow run; failures are loud.
• mode: check reports drift and writes nothing.
• Apply reads first and writes only what differs: a converged repository gets zero writes, and a second apply is a no-op.
• Multi-repo mode manages a whole fleet from one admin repository.

Setup and the full section reference: README.

1.0.0 (2026-07-22)

Features

• actionable errors, per-call debug tracing, and coverage docs (fe9c9c5)
• add discovery filters for multi-repo “*” mode (1d6531f)
• adopt octokit, actions/core, and zod for transport, IO, and validation (ff89bb6)
• api-version input, self-updating pre-commit, bundle-freshness test (a836718)
• apply own settings with the action at HEAD (4dac8fc)
• declarative section permissions and endpoint dictionaries (de24164)
• five new settings surfaces, audit fixes, and structural refactors (30e2dd2)
• forward-compatible key routing in the actions section (1818569)
• full passthrough in every section plus coverage inventory (34f108a)
• initial settings-as-code action (6e4857f)
• multi-repo mode with central files, remote settings, and a defaults layer (04b379e)
• node24 runtime and husky pre-commit hook (ba04830)
• preflight barrier makes strict applies all-or-nothing (a92173f)
• publish generated settings.yml JSON Schema (b706fa9)

Bug Fixes

• enforce read-only preflight probes and guard check-mode purity (009def9)
• environments PUT status and write-throttle scaling, found by the new e2e fuzz harness (b032024)
• escape backslashes before pipes in the summary table (6684569)
• format the e2e mock files that landed mid-refinement (8911068)
• make the unrecognized actions-key note mode-aware and name the enabled value (1d3bc0a)
• pin bun via .bun-version so CI rebuilds the bundle byte-identically (4e7f2bc)
• print the final result on stdout (76b258d)
• rate-limit discovery advice, shared constants, docs pinned to code (cf8f291)
• reject duplicate ruleset and branch declarations before any API call (441ed49)
• shape-check the fields section handlers dereference (c9a8585)
• teams org grading, nightly issue auto-assignment, and fuzz artifact hygiene (f0378f0)
• unique marketplace name and shorter description (9508134)

Full Changelog: https://github.com/Vivswan/repo-settings-as-code/commits/v1.0.0

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates Rust script execution by interpreting a practical subset of the language without compiling the entire project. It supports running scripts directly as binaries or executing snippets in compiled form, validating without running, and providing a list of supported methods per receiver and engine. The action is particularly useful for small scripts that do not require compilation to run efficiently.

What’s Changed

Full Changelog: https://github.com/VladasZ/rustscript/compare/v0.1...v0.1.9

• This action is used across all versions by 1 repositories.

Action Type

This is a Node action using Node version 24.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action vmactions/hurd-vm allows users to run CI tests in the Hurd operating system. It automates the setup and execution of CI scripts on Hurd, a Unix-like operating system derived from GNU/Hurd. This action provides features such as sharing code between the host and VM, using different synchronization methods like rsync, nfs, or scp, and NATting ports to allow for remote access to the VM.

What’s Changed

init

Full Changelog: https://github.com/vmactions/hurd-vm/commits/v1.0.0

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action checkout checks out a repository in treeless mode, which includes only the files and directories specified in .gitattributes. This is useful for creating lightweight clones of repositories that do not include unnecessary binary or large files. The action automates the checkout process by using the default ref (github.ref) if not provided, and supports specifying the repository name and server URL as inputs.

What’s Changed

Full Changelog: https://github.com/void-musl/checkout/commits/v1.0.0-treeless

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action automates the process of uploading files to a specific release on GitHub. It simplifies the task by requiring only essential inputs such as the tag, release name, file pattern, and GitHub token. The main functionality is to allow developers to easily publish build artifacts or binary files directly from their workflows.

What’s Changed

Full Changelog: https://github.com/void-musl/upload-release/compare/v1.0.1...v1.0.2

• This action is used across all versions by ? repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

The GitHub Action wartzar-bee/ci-guardrail calculates the cost difference between the base and head branches of a PR, posts a comment with the delta and top contributors, and can block the build if the token cost increases beyond a configured threshold. It uses @wartzar-bee/tokenscope to estimate token costs based on file size and complexity, providing insights into potential cost regressions.

What’s Changed

v1.1.1 — self-test CI + tests badge

Patch release. No change to the action’s behavior (entrypoint.sh / action.yml untouched vs v1.1.0). Backward compatible — uses: wartzar-bee/ci-guardrail@v1 picks it up when the v1 tag is moved.

Added

• Self-test CI. .github/workflows/test.yml runs the 50-test integration suite (bash test/run-test.sh) on every push to main and every pull_request. The suite mocks tokenscope and git on PATH, so it needs only bash + python3 (both preinstalled on ubuntu-latest) — no npm/tokenscope install.
• Tests badge in the README (above the Marketplace badge). For a GitHub Action, a visible green self-test is a direct trust/conversion signal for the repo visitors the dev.to announcement + Marketplace listing drive here.

Fixed

• README hero-line overclaim (honesty fix). The top-of-README tagline claimed the action “comments on the responsible lines”. It does not — it comments on the responsible files (tokenscope scan emits file-level token data with no line-level attribution; the PR comment renders per-file delta tables). Corrected “lines”->“files” so the flagship’s first-impression line matches action.yml’s Marketplace description and the actual behavior. Brand HARD RULE: never claim a capability the product lacks.
• Publish allowlist gap. .publish-include now lists .github/workflows/test.yml. It previously did not, so the tests badge added in the prior tick would have pointed at a workflow that never shipped to the public repo — a broken (404) badge. Now the workflow ships, the badge runs, and it goes green after the first push.

Tests: 50/50 pass. Pre-publish audit: CLEAN.

• This action is used across all versions by 0 repositories.

Action Type

This is a Composite action.

Go to the GitHub Marketplace to find the latest changes.

Action Summary

This GitHub Action creates a headless, scriptable, CI-ready test harness for testing Claude Cowork skills. It reproduces Claude Cowork’s observable runtime contract closely enough to test the skills you write across multiple scenarios and in CI environments, without using the locked Desktop app. The action supports various fidelity tiers including replay, lint, container, hostloop, and microvm, each requiring different prerequisites such as a running agent, a Claude token, and a Docker or Lima runtime.

What’s Changed

Added

• migrate-run-dir — convert pre-layout run dirs to the per-turn turns/<N>/ layout, in place. A run dir written before the per-turn layout keeps result.json / run.jsonl / trace.json / resources.jsonl at its root. Once the legacy read layer is removed, those dirs become unreadable to verify-run / diff / inspect / stats; this command converts them so the history survives the change instead of having to be re-run.

  Dry-run by default — --write applies, and --scenario <name> scopes the run to a single scenario so a rollout can be staged: migrate one, verify it, then do the rest. It renames rather than copies, so file mtimes (the recency signal stats and status --latest-for rank by) survive untouched, and it restores directory mtimes afterwards. An interrupted run records a journal outside the run dir and is finished by re-running the command. Anything it cannot resolve unambiguously — a root artifact that is neither a duplicate nor placeable, telemetry whose turn boundary cannot be dated or that spans more than two turns or whose samples would land in a turn no transcript or result evidences, a dir with no transcript at all — is refused and named. The one inference it makes is positional: an EMPTY file has no content to attribute, so it follows its position to an EVIDENCED turn (its own, by name or by rootArtifactTurn when a root transcript or result exists to move there) — never one it would mint. The same never-mint rule holds on the content path: a stray resources.turn-N.jsonl cannot manufacture turns/N/ whether it is empty or carries samples, and a fully-archived dir’s trailing telemetry cannot manufacture the next turn out of arithmetic alone. Exit 1 when anything was refused, so a CI caller sees unfinished work. After a --write that migrated or recovered anything, it prints a reminder to rebuild the index (cowork-harness stats --reindex), since the index keys a row’s timestamp off result.json’s mtime and the files have moved.

• prune skips scenarios with a migration in flight. Between an interrupted migration and its recovery a run dir’s mtime reflects the migration, not the run — and prune ranks keep-slots by that mtime, so it could evict a newer run in favour of a half-migrated older one. It now defers those scenarios and says so.

• critique surfaces the GRADED turn’s outcome and skillHash in its own report (gradedOutcome / gradedSkillHash in JSON, and in the text header), and writes the graded result under the stable name result.graded.json. critique runs two turns into one run directory, so after the resume result.json is the reflection turn’s and the graded turn is archived as result.turn-1.json — the correct file to read was the lower number, the opposite of every other multi-run convention. A harvester reading result.json silently ingested the reflection turn’s numbers: valid-looking, wrong, and unsignalled. Reported by a consumer building exactly that harvester; a documentation-only fix would have helped only readers who already knew to look.

• exec_infra_error verdict signal (WARN) — a container exec that failed for infrastructure reasons, as distinct from the fail-severity infra_error (a supervising process died). One failed command no longer contaminates a whole run’s evidence.

• RunResult.infraErrors[].source is now an enum — hostloop-sidecar / hostloop-exec / egress-sidecar. The origin is what drives severity, and it is carried through the frozen cassette so replay reaches the same verdict as the live run.

• Capability use-scan health — an unreadable or partially unparseable events.jsonl is now reported as a degraded scan instead of being indistinguishable from a complete scan that found nothing.

• Every critique limitation is now tagged with WHY it exists, not just what it is — structural (permanent, architect around it), unverified (unproven, not known-impossible), deliberate (a design choice), not-built (simply absent). The distinction a reader needs is rarely “what can’t it do” but “should I design around this forever, or wait for it?” Container-tier-only is unverified: the resume-continuity proof was run against the container tier’s Linux ELF, and hostloop runs a different (native) agent binary, so the proof does not transfer — nothing suggests hostloop would fail, nobody has run it. Lifting the pin needs BOTH a live resume-continuity proof at hostloop against its native binary AND the follow-on work that proof unblocks (unpinning three hard-coded container sites, stamping the tier on the session manifest so a cross-tier resume fails loud, and plumbing host-write consent) — evidence alone is not sufficient. A consumer read that pin as permanent and built a second test lane around it. The tags appear in critique --help and docs/critique.md, generated from one source.

• critique --help’s KNOWN LIMITATIONS block is generated from that source, and CI asserts the shipped binary’s output, the docs bullets, and their tags all agree.

Changed

• ⚠️ BREAKING: per-turn run-directory layout, single shape — the root-level result.json compatibility copy is REMOVED. A run directory that holds several turns (any --resume, and every critique) writes each turn’s result.json, run.jsonl, trace.json and resources.jsonl into turns/<N>/, once, under its final name — nothing is renamed or overwritten as later turns arrive. chat now goes through the same layout too (always turns/1/ — a chat session mints a fresh dir per invocation and never resumes). <outDir>/result.json no longer exists — there is no root compat copy of any per-turn artifact, on ANY run dir. Read turns/<N>/result.json directly (turns/1/ for a single-turn run), or — for critique — the unchanged result.graded.json / trace.graded.json role aliases. Cumulative streams (events.jsonl, timeline.jsonl) and session state are unchanged, so critique’s byte-offset turn-isolation proof and cassette capture are unaffected.

  Two prior shapes are now REFUSED, loudly, by name, instead of being silently misread:

  • a pre-layout run dir (written before turns/<N>/ existed: root result.json/run.jsonl, or a name-mangled result.turn-<N>.json archive, no turns/);
  • a mixed run dir (a pre-layout dir resumed under CURRENT code before this release — turns/ present and a stray root/archived file).

  verify-run, inspect, scaffold, diff, status --latest-for, and a resumed --session-id all refuse these with a message naming the shape found and pointing at trace <dir> — which still works fully, since every one of its views derives from events.jsonl, which never moves. stats --reindex counts them as skipped and names the remedy rather than dropping them from the index quietly.

  Migration: cowork-harness migrate-run-dir converts a pre-layout dir in place (dry-run by default), preserving the file timestamps stats and status --latest-for rank by. diff and status --latest-for are called out because their pre-refusal behaviour was the dangerous kind: diff reported two genuinely different runs as identical and exited 0, and status --latest-for could select a different run than the newest and report its verdict — a CI script reading .verdict.pass got a green light for a red run.

  Previously the latest turn lived at the root while earlier ones were name-mangled archives, so a file’s name depended on whether a later turn ever happened; that shape produced a wrong-turn read, a destroyed trace, and a dropped index row — this release’s read-side (turnArtifactPath / listTurns in turn-layout.ts, with the old readTurnResult deleted for having zero production callers) no longer has a legacy-resolving branch at all, so that class of bug is now unrepresentable rather than merely fixed. The Python SDK’s _latest_run_jsonl likewise now raises loudly on a pre-layout dir instead of silently falling back to a root run.jsonl that (for any current-layout dir) is a path to nowhere.

• Platform baseline synced to Desktop 1.24012.1 (baselines/desktop-1.24012.1.json, now what baseline: latest resolves to). The staged agent binary is 2.1.217 (native app + VM ELF, new sha256 for each). The baseline moved in two steps this release — an earlier sync to 1.24012.0 (agent 2.1.215), then to 1.24012.1 — with no prompt, spawn-env, or egress-allowlist drift across either: spawn.env is byte-identical to 1.22209.3, the same 15-domain allowlist and gvisor mode carry over (the effort map is the one spawn field that changed — see the sonnet-5 delta below), and the deriveSpawnEnv / checkSpawnContractFacts oracles stay green against the live asar. The substantive deltas all came from the 1.24012.0 step and carry forward unchanged: claude-sonnet-5 joins the per-model effort map (low|medium|high|xhigh|max, recommended medium, modes auto); the coworkRuntimeConfig gate drops its pluginsFullSyncStalenessMs key (never modeled here, inert); and the dormant autoModeOverridesAlwaysAllow sentinel fired — see below. 1.24012.1 itself adds only the agent bump: the 2.1.217 binary can emit the VCS SDK events code_change_published / vcs_state_changed (SDK floor 2.1.216), which the harness surfaces as a system_event (its existing graceful degradation of an unmodeled system event, unchanged), and the binary’s native skill-discovery enable predicate widened to three branches — still inert here, since real Cowork’s model-visible surface is the Desktop SDK-MCP discovery servers, not the native tools. The example cassettes' fingerprint.baseline tracks the new baseline.

• The autoModeOverridesAlwaysAllow gate (4200321681) flipped absent → on (source: force) and was revisited as its pin intended. It stays unmodeled, deliberately: binary-verified in 1.24012.0, both call sites only override an already-existing always-allow decision — the session rule cache (approvedToolNames) and scheduled-task auto-approval — each further gated on permissionMode and isDestructiveConnectorTool. The harness persists neither, so it already prompts wherever the gate makes Cowork prompt; enabling it moves real Cowork toward harness behavior rather than away. Revisit only if the harness gains a persistent per-tool approval cache.

• The staged agent (2.1.217) enforces sub-agent fan-out caps, so dispatch_count_max is now framed as a budget UNDER Cowork’s cap, not a reproduction of it. Because the harness spawns the real binary, a run that fans out past the agent’s caps now errors from the binary itself: a concurrent cap (CLAUDE_CODE_MAX_CONCURRENT_SUBAGENTS, default 20; error subagent_concurrency_cap, new in 2.1.217) and a per-session cap (CLAUDE_CODE_MAX_SUBAGENTS_PER_SESSION, default 200; error subagent_count_cap, present since ≤2.1.215). The scenario schema’s dispatch_count_max description, the assert over-budget message, and SPEC §10 no longer claim “Cowork imposes no in-conversation Task-dispatch cap” — that claim was stale. The harness does not reproduce the caps; it inherits them by running the binary.

• A host-loop exec infrastructure failure now WARNS instead of failing the run. ⚠️ Upgrade note: a run that previously exited 1 because one docker exec failed will now exit 0. A dead sidecar still hard-fails. Known residual, documented in docs/scenario.md: if every exec failed the agent ran nothing and the run still only warns — inspect result.infraErrors when a run looks suspiciously empty.

• A model-requested bash timeout_ms expiry is no longer classified as an infrastructure failure. The model now receives its command’s own partial output with Command timed out after <duration> merged into stderr — matching real Cowork, verified against the staged agent binary — instead of an opaque [infrastructure error: see run log for details].

• The agent’s spawn env now always carries a normalized IANA TZ — matching Desktop, which injects Intl.DateTimeFormat().resolvedOptions().timeZone unconditionally. Previously TZ was forwarded only when the host shell exported it, and forwarded raw, so a host with no TZ set — or a legacy/non-IANA export (US/Eastern, EST5EDT) — diverged from real Cowork’s date/“today” rendering inside the agent.

• The tool_available assertion now names its evidence limit. It evaluates against the run’s eagerly-loaded tool set (the SDK init manifest in result.json); a factory-deferred tool — e.g. the skill-discovery MCP tools, loaded on demand via a ToolSearch round-trip — can be genuinely available in the run yet miss here, a false negative. The assertion still fails on a miss; the failure message now states that eagerly-loaded scope rather than implying provable unavailability.

• An explicitly requested --dotenv file now fails loud. ⚠️ Upgrade note: an unreadable file, or a path that is a directory, previously fell through to lower-precedence .env sources while still printing a success line — so a typo’d path silently ran against the wrong credentials. It is now a usage error. Automatic .env discovery is unchanged (still best-effort).

• diff no longer reports identical when only one side has an artifact manifest. ⚠️ Upgrade note: such a comparison previously exited 0; it now exits 1, because unavailable evidence is not evidence of equality. Both-sides-missing still does not veto identity. The --output-format json envelope gained an artifactsAvailability key.

• stats --reindex merges rows by per-completion identity (outDir + a new turn field) rather than by outDir alone, and reports rejected symlinked run directories.

Fixed

• verify-run now REFUSES a multi-turn run directory instead of certifying the wrong turn. Root result.json is the latest turn; on a critique directory that is the reflection turn while the scenario describes the graded one. Previously the cumulative gate scan false-FAILED on the other turn’s gates — wrong, but loud. The refusal names result.graded.json / turns/1/result.json so the caller can still reach the graded turn.

• trace no longer mixes turn scopes. After timeline reads became turn-scoped, --view tool-durations showed the latest turn while the tools/questions/dispatches views still showed every turn — two views of one run directory describing different scopes. All views are now the latest turn, and a ::notice:: reports when earlier turns exist rather than hiding them. Its cache-read footer and gate-provenance (answeredBy) views also now say when a result is not turn-addressable (a pre-layout dir) instead of silently omitting the cache-read ratio / labels.

• prune no longer demotes an unmigrated pre-layout run dir to the junk tier. Its real-run predicate keyed on hasTurnDirs || events.jsonl, reasoning only about what current writers produce — but prune ranks history, including the legacy dirs migrate-run-dir exists to preserve, so a legacy dir with no events.jsonl could be evicted ahead of an empty scaffold. It now also counts a legacy / mixed shape as a real run. (Distinct from the in-flight-migration deferral above — this is about which dirs count as real at all.)

• A resumed turn was judged on the PRIOR turn’s evidence — three wrong-verdict paths. events.jsonl is append-only across turns with no per-turn marker, and three whole-file scanners decide a run’s outcome: scanEvents (outputs-delete / host-path-leak → fail signals, and an authored no_delete_in_outputs), findUngatedPathToolCalls (→ a run-level error at hostloop), and detectCapabilityUse (→ missing_capability, a fail signal, which fires on the default lean image). So on any --resume — and every critique reflection turn — turn 1’s delete, gated tool call, or capability use FAILED turn 2. A turn-start marker now scopes all three to the current turn. resources.jsonl had the same shape (turn 1’s peak RSS judged against turn 2’s max_peak_rss_bytes) and is archived per turn. Single-turn runs write no marker, so their events.jsonl is byte-identical and no cassette is affected. Missing marker ⇒ whole-file scan, i.e. fail-closed.

• A resumed turn’s telemetry included the PRIOR turn’s events, and could produce a false PASS. timeline.jsonl is append-mode with a fresh header per turn, but readTimeline returned every line after the first as an event — so on any --resume (and every critique reflection turn) the current turn’s toolDurations/skillActivity/subagents folded in the previous turn’s tool calls. Because the skill_tool_used assertion evaluates against that same skillActivity, a turn-1 skill window could satisfy a turn-2 assertion. The reader now returns only the current turn’s segment. The file stays one append-only stream, so critique’s byte-offset turn-isolation proof is unaffected.

• A resumed turn destroyed the prior turn’s trace.json. Because it is rebuilt and overwritten on every completion, the earlier turn’s trace was deleted rather than preserved, so a critique lost the graded turn’s trace entirely. Each turn now owns its own turns/<N>/trace.json, written once and never overwritten, and critique additionally writes trace.graded.json beside result.graded.json.

• stats --reindex dropped every non-latest turn when rebuilding from the runs tree. It read only the root result.json per run directory, so a resumed session’s earlier turns vanished — and on a critique directory the root file is the reflection turn, so it was the graded rows that were lost. Every turn under turns/<N>/ is now indexed as its own completion; result.graded.json — a root-level copy of the graded turn — is deliberately not matched, so it cannot double-count. A dir that has not been migrated is counted as skippedLegacy and reported with the remedy, never dropped silently.

• An ambient GIT_DIR silently computed the wrong skill file set. Git hooks export GIT_DIR (and GIT_INDEX_FILE) into every child process, and with GIT_DIR set but no GIT_WORK_TREE git stops inferring the work tree from cwd and treats cwd as the repo root. gitTrackedSet’s rev-parse --show-toplevel probe therefore still succeeded — so the not-a-repo raw-walk fallback never fired — while git ls-files -- . returned the entire repo index as root-relative paths instead of the directory-relative ones. Measured on this repo: 2 tracked files became 625 wrong ones. That set feeds both skillHash and the mount-copy filter, so any run invoked from a git hook (or from CI that exports GIT_DIR) got a wrong hash and a mount filter pointed at paths that do not exist under the skill dir. The visible symptom was the repo’s own pre-commit hook reporting committed example cassettes as [stale] skill files changed since record on every parity sync. skillCommit had the same defect: git -C <dir> is overridden by an ambient GIT_DIR, so every skill dir resolved to that foreign repo’s HEAD — recording a foreign commit as the skill’s provenance and masking dirs that are genuinely in different repos. Both call sites now spawn git with GIT_DIR / GIT_WORK_TREE / GIT_INDEX_FILE stripped, via one shared helper so they cannot drift. run-index’s gitInfo and doctor’s worktree probe deliberately keep inheriting — they are asking about the ambient repo.

• stats --reindex destroyed multi-turn history. Every --resume turn — and critique’s task + reflection pair — writes to one outDir, so keying by directory collapsed N completions into one, silently changing run counts, pass rates and costs.

• Host-loop sidecar failures never reached the verdict. They were appended straight to events.jsonl, which no live drive re-reads, so a dying sidecar left the run green; a signal-only termination (OOM, SIGKILL) was recorded nowhere at all.

• result.json was written non-atomically at all three producers, so an interrupted write could leave the canonical record truncated.

• Corrupt index rows were blind-cast, letting one malformed row crash stats or fabricate a pass/cost value; reindex also followed symlinks out of the runs root.

• scaffold turned unavailable artifact evidence into “no artifacts”, permanently encoding a false “this run produced nothing” claim into a generated scenario.

• critique treated a vanished turn-1 evidence file as genuinely empty evidence rather than an integrity failure. A stream that was legitimately zero bytes at capture is still reported clean.

• critique’s exit-code table omitted 1. Exit 1 is reachable on operator interrupt (SIGINT/SIGTERM); a sweep wrapper treating it as impossible misreads a cancelled run as a crash.

• Documented that after critique’s resume, result.json is the reflection turn’s result — the graded turn is archived as result.turn-1.json. Reading the wrong one yields a valid-looking wrong number.