Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.5.0. 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 an unofficial test harness designed to emulate Claude Cowork’s observable runtime contract. It allows developers to script and automate the testing of skills across various scenarios in a headless, CI-friendly environment. The action supports multiple fidelity levels (e.g., replay, full) to capture detailed information about what the agent did, not just its behavior. This helps ensure that skills work as expected within the constraints defined by Cowork’s runtime contract.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.4.0. 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 cowork-harness is an unofficial tool that creates a scriptable and CI-friendly test harness to emulate the observable runtime contract of Claude Cowork. It helps developers test their skills in multiple scenarios headlessly, across many CI jobs, without relying on the locked Desktop app. The action automates the reproduction of constraints such as sealed filesystem, default-deny egress, and MCP-only cross-boundary, providing evidence of what the agent actually did during a run. This is particularly useful for ensuring that skills are compatible with Cowork’s environment, even if they don’t pass strict claude -p runs.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.3.0. 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 The cowork-harness action is a test harness for Claude Cowork skills that emulates the observable runtime contract closely enough to test them across multiple scenarios and in CI environments. It reproduces constraints like the sealed filesystem, default-deny egress, and MCP-only cross-boundary, providing evidence of what the agent actually did rather than just what it said. This helps in ensuring skill reliability without relying on a locked-down Desktop app and is not a guarantee but a strong signal for test outcomes.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.2.0. 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 The cowork-harness action is a scriptable and CI-ready tool that reproduces Claude Cowork’s observable runtime contract to test agent skills in headless environments across multiple scenarios. It provides evidence of what an agent actually did, not just what it said it did, by recording every run. This helps ensure that the tested skills are robust against constraints like the sealed filesystem, default-deny egress, and MCP-only cross-boundary access.

Version updated for https://github.com/yaniv-golan/cowork-harness to version 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 The cowork-harness GitHub Action is a scriptable and CI-friendly test harness that reproduces Claude Cowork’s observable runtime contract closely enough to test skills without using the locked Desktop app. It emulates constraints such as sealed filesystem, default-deny egress, MCP-only cross-boundary, and provides detailed evidence of agent behavior through recorded runs, enhancing test reliability beyond bare CLI execution.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.0.1. 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 cowork-harness action automates the testing of Claude Cowork skills by providing a scriptable and CI-friendly test harness that closely reproduces the observable runtime contract. It ensures that tests clear the constraints that break skills in Cowork, offering evidence of what the agent actually did during a run, not just what it said it did. The tool supports various fidelity tiers and provides an interface to inspect and analyze test results.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v2.0.0. 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 The cowork-harness action is a scriptable and CI-friendly test harness that replicates the observable runtime contract of Claude Cowork, enabling developers to thoroughly test their skills across various scenarios without using the locked Desktop app. It captures evidence of what an agent actually did during a run, including invoked skills, files read, network requests, and shown options. The action supports different fidelity tiers for different testing needs, from headless operations to full-fledged simulations in Docker or Lima environments.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.25.0. 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, cowork-harness, provides an automated test harness that reproduces Claude Cowork’s runtime contract closely enough to test skills. It allows developers to run and test their skills across various scenarios without the need for a locked-down Desktop app, ensuring that tests are accurate and reliable. The action supports multiple fidelity tiers, including free demos, global installs, linting with Python, and live tiers that require specific resources such as Claude Desktop and a token.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.23.0. 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, cowork-harness, is a test harness designed for testing Claude Cowork skills. It automates the reproduction of the observable runtime contract closely enough to run tests headless and in CI environments across various scenarios. The key capabilities include reproducing both behavior and limitations such as sealed filesystem and default-deny egress. It provides a green test result when all conditions are met, ensuring that real Cowork runs would also pass.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.22.0. 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 cowork-harness is a testing tool that automates the testing of Claude Cowork skills by reproducing its observable runtime contract, including limits such as sealed filesystem and default-deny egress. It allows developers to script and automate end-to-end tests across various scenarios in a headless manner, making it suitable for CI pipelines without relying on the locked-down Desktop app. The tool provides options to test local skills with different fidelity levels (protocol, replay, live), each requiring varying prerequisites including token, runtime setup, and agent configuration.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.21.1. 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 This GitHub Action is a tool designed to automate and test Claude Cowork skills without using the actual Desktop application. It mimics the runtime environment closely, allowing developers to run tests across various scenarios in a headless manner, including CI pipelines. The action supports both protocol-only and live test tiers, requiring different setups for token-based or desktop-based testing.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.20.0. 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 The cowork-harness action is a scriptable, CI-friendly test harness that reproduces Claude Cowork’s observable runtime contract closely enough to test skills without the locked Desktop app. It creates a sandboxed environment, simulates the agent and its constraints (e.g., sealed filesystem, MCP-only cross-boundary), and can run tests in various fidelity tiers such as replay or live with different requirements like token-free or with real resources.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.17.0. 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 This GitHub Action is designed to automate and test Claude Cowork skills in a scriptable, CI-ready manner without relying on the Desktop app. It reproduces key aspects of Cowork’s runtime contract, including sealed filesystem, default-deny egress, and MCP-only cross-boundary behavior. This allows developers to test their skills across various scenarios and ensure compatibility with the platform before deploying them.

Version updated for https://github.com/yaniv-golan/cowork-harness to version v1.16.0. 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 action is a scriptable, CI-friendly test harness that emulates Claude Cowork’s observable runtime contract to test skills without using the locked Desktop app. It reproduces behavior and limitations such as the sealed filesystem, default-deny egress, MCP-only cross-boundary access, and can run in headless mode across various scenarios and CI jobs. The action is useful for developers to ensure their skills work as expected and identify any potential issues before deployment.