The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a bundled Python script, scripts/collect-release-evidence.py, to perform data collection and verification tasks. The script uses subprocess.run to invoke the GitHub CLI (gh) for API interactions. This is a functional requirement for the skill's operation.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external content, such as pull request titles, bodies, and commit messages, which are fetched from GitHub repositories. This represents a standard injection surface for agents processing text.
Ingestion points: Data is ingested through gh api calls in scripts/collect-release-evidence.py and via Git CLI commands in SKILL.md.
Boundary markers: The instructions explicitly warn the agent to treat PR bodies as context rather than direct evidence and to verify all claims against the final code diff. It also uses a dedicated verify command in the bundled script to enforce structural and source integrity.
Capability inventory: The skill has the capability to execute the bundled Python script, the gh CLI, and the git CLI. It performs file system operations (writing evidence and draft notes) strictly within the operating system's temporary directory to avoid polluting the workspace.
Sanitization: The bundled script validates repository names using regular expressions and uses urllib.parse.quote to sanitize tag and branch references before including them in API endpoints.

git-remote-release