The Agent Skills Directory

[COMMAND_EXECUTION]: Executes repository management commands using git, gh (GitHub CLI), and jq. Write operations such as gh pr create and git push are gated by mandatory user confirmation.
[DATA_EXPOSURE]: Implements a proactive security guardrail to scan for sensitive information like secrets, tokens, and .env files in commit messages and diffs to prevent accidental exposure in pull request bodies.
[PROMPT_INJECTION]: Identifies an indirect prompt injection surface as the skill ingests untrusted data from commit messages, diffs, and pull request templates to generate summaries.
Ingestion points: Commit messages via git log, diffs via git diff, and repository PR templates (SKILL.md, REFERENCE.md).
Boundary markers: The skill explicitly presents a 'PR plan' to the user for review and confirmation before any remote operations or PR creations are performed (SKILL.md).
Capability inventory: Utilizes git and gh for repository state analysis and network-based PR management (REFERENCE.md).
Sanitization: Contains instructions to detect and exclude secrets or credentials from generated PR content (SKILL.md).

pull-request