Version updated for https://github.com/Zandereins/schliff to version v8.7.0. 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 This GitHub Action automates the scoring of AGENTS.md files using the Schliff tool. It ensures that AI instruction files remain consistent across different systems by providing deterministic quality scores based on a versioned rubric. The action helps detect and prevent issues in instruction files that could degrade LLMs, ensuring better reliability in AI development workflows.

Version updated for https://github.com/Zandereins/schliff to version v8.6.1. 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 This GitHub Action automates the scoring of AGENTS.md files to ensure consistent quality across different environments. It uses a deterministic rule engine to evaluate instruction files against explicit rubrics, preventing degradation due to inconsistencies or rotting prompts. The action ensures that AI tools are consistently evaluated and can be trusted for their accuracy and reliability.

Version updated for https://github.com/Zandereins/schliff to version v8.6.0. 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 This GitHub Action automates the scoring of AGENTS.md files using Schliff, a tool that provides deterministic quality scores based on a versioned rubric. It helps ensure that AI instruction files remain consistent across different environments and tools by comparing them against an explicit evaluation standard. The action is designed to help identify and address issues in AI instruction files before they degrade over time, ensuring better performance and reliability of AI-driven tools.

Version updated for https://github.com/Zandereins/schliff to version v8.5.0. 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. What’s Changed Fixed verify scores under the detected format profile (#102, closes #101): schliff verify AGENTS.md no longer fails spuriously (was 27.7/F under the SKILL profile on a file score grades 91.6/A). SKILL.md verdicts unchanged. Positional negation in operational_coverage (#96): contrastive sentences (“run X, never Y directly”) keep the recommended command. Dead-marker detector matches marker tokens, not English prose (#97, closes #93). Corpus goldens re-derived. Security Runtime scorer prompt nonce-hardened (#99) — defense-in-depth, dimension remains opt-in. Added Theme-aware README hero + social-preview asset (#106); star-notify workflow (#98); grouped Dependabot action bumps (#107). Docs README redesigned (#100): AGENTS.md-first, every number ground-truthed against the released engine, honesty-hardened. Full changelog: https://github.com/Zandereins/schliff/blob/main/CHANGELOG.md

Version updated for https://github.com/Zandereins/schliff to version v8.4.0. 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. What’s Changed Added operational_coverage dimension for AGENTS.md (#83). Measures whether an AGENTS.md actually equips a coding agent to operate the repo: real setup/build/test commands (command-family classification, doc-wide, headings never gate) plus code-style / gotchas / PR directive sections with concrete code tokens. Surfaced in the CLI dimension table, the GitHub Action’s PR comment, and accepted by the leaderboard submit API. Changed BREAKING (scores): the AGENTS.md headline profile is now structure 0.40 / operational_coverage 0.40 / efficiency 0.20 (was 0.5/0.5). efficiency was a validated gameable proxy — a junk-fence-stuffed doc scored 92.5/A while the same real commands written inline scored 70.0/C. All AGENTS.md scores re-baseline (30-file corpus: mean 61.06, no file reaches S). SKILL.md / CLAUDE.md / .cursorrules / system-prompt scoring is byte-identical to 8.3.0. Security Fixed a ReDoS in the operational_coverage heading regex (quadratic on whitespace-only heading lines) before it ever shipped — found by a 75-agent adversarial review pass, together with a directive-gate gaming hole, a fence-state desync, and 12 command-recall bugs. Full record: docs/specs/agents-md-operational-coverage.md §11. Full changelog: https://github.com/Zandereins/schliff/compare/v8.3.0...v8.4.0