The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill utilizes standard development tools (git, npm, gh) for their intended purposes.
[DATA_EXFILTRATION] (SAFE): Sensitive operations like git push are restricted to branch updates, and no external network calls to untrusted domains are present.
[INDIRECT_PROMPT_INJECTION] (LOW): The skill reads external PR templates which could contain malicious instructions.
Ingestion points: Reads files from .github/pull_request_template.md and related paths.
Boundary markers: Absent for template content.
Capability inventory: git commit, git push, gh pr create, and npm run preflight execution in SKILL.md.
Sanitization: Implements high-quality sanitization by using --body-file for the GitHub CLI to avoid shell escaping vulnerabilities.

pr-creator