The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes Bash to perform repository management tasks through git and the GitHub CLI (gh). It executes commands to list PRs, compare branch histories, and update PR titles and bodies. These operations are restricted to the repository context and are necessary for the skill's primary function.
[INDIRECT_PROMPT_INJECTION]: The skill is exposed to indirect injection via commit messages and external PR data. It proactively addresses this by implementing a mandatory 'Fact-check' stage (Step 4.5). This step invokes a separate agent to validate that every claim in the draft description is grounded in the actual file changes (diffs), ensuring that the agent does not blindly trust or execute instructions embedded in commit logs.
[EXTERNAL_DOWNLOADS]: The skill instructions reference a remote repository for testing fixtures at https://github.com/ashe-li/agent-skills/tree/main/release-pr/fixtures. This is a vendor-owned resource (authored by ashe-li) and is used to verify the skill's own performance and coverage against known failure modes.
[DYNAMIC_EXECUTION]: The skill employs modular design by invoking the evidence-gate skill at runtime to handle specific validation logic. This is a standard platform capability used here to enforce data integrity and security checks.

release-pr