The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes standard command-line tools such as git and gh (GitHub CLI) to manage repository branches, diffs, and pull requests. These operations are essential to the skill's functionality and are performed within the local repository context.
[INDIRECT_PROMPT_INJECTION]: The skill reads repository data (commits, diffs, and templates) to populate PR fields. While this creates a surface for indirect prompt injection where malicious repository content could attempt to influence the agent's PR drafting, the risk is negligible as the output is restricted to the PR description and is intended for human review on GitHub.
Ingestion points: .github/pull_request_template.md, git commit messages, and repository diffs.
Boundary markers: The skill uses a structured markdown template to separate sections.
Capability inventory: git push and gh pr create in SKILL.md.
Sanitization: None specified for interpolated text.
[SAFE]: The skill provides a --safe flag that invokes an external security tool (safe-repo) to check for secrets and sensitive data before pushing code, which follows security best practices.

to-pr