The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to execute shell commands specified in a local release.config.json file. These commands are used for 'readiness gates' (e.g., linting, testing, and production builds). The execution is performed by the agent at the user's direction based on the provided configuration.
[COMMAND_EXECUTION]: The workflow involves multiple Git operations, including git checkout, git log, git tag, and git push. These are standard requirements for a release management tool.
[DATA_EXPOSURE_&_EXFILTRATION]: No evidence of hardcoded credentials, sensitive file exfiltration, or unauthorized network requests was found. The Python engine (scripts/release.py) uses only standard library modules and performs file system operations limited to project configuration and version files.
[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from Git commit history and external standards documentation.
Ingestion points: Commit subjects via git log, project configuration files (release.config.json), and the COMPATIBILITY.md document.
Boundary markers: The skill explicitly instructs the agent to 'hand-curate' changelog entries and perform a manual 'Compatibility review', which acts as a human-in-the-loop check against unintended instruction obedience.
Capability inventory: The agent has the capability to write to the file system (version files, changelogs), execute shell commands defined in release.config.json, and perform Git operations (commit, tag, push).
Sanitization: The Python engine uses targeted regular expressions and atomic string substitution to ensure version file updates are precise and do not corrupt adjacent data.

release