The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes common CLI tools including git, gh (GitHub CLI), and npm to perform repository management, build artifacts, and interact with GitHub Actions.
[EXTERNAL_DOWNLOADS]: The version publishing workflow uses npx bumpp, which dynamically downloads and executes a package from the npm registry to handle version incrementing.
[SAFE]: The skill enforces data safety by explicitly instructing the agent to avoid committing sensitive information or secrets during the git workflow.
[SAFE]: Security is enhanced through mandatory interactive confirmation steps (Step 4 in releasenote.md and Step 3 in version_publish_main.md), ensuring that no changes are pushed or published without explicit user review.
[SAFE]: The GitHub Action failure triage workflow uses isolated git worktrees (git worktree add), which prevents accidental modification of the main development branch and ensures that experimental fixes are contained.
[SAFE]: The potential for indirect prompt injection from processing external data (such as git logs or GitHub Actions failure logs) is mitigated by the workflow design, which requires a human-in-the-loop to verify findings and approve any generated PRs or release notes.
Ingestion points: git commit logs and GitHub Action failure logs (gh run view --log-failed).
Boundary markers: None explicitly defined in the command templates.
Capability inventory: Git branch management, code modification, pushing to remote, and PR creation.
Sanitization: All generated content and proposed code fixes are subject to mandatory user review before execution.

git-workflows