repository-cleanup
Pass
Audited by Gen Agent Trust Hub on Aug 31, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses local Git commands (git branch, git stash, git worktree) and the GitHub CLI (gh pr list) to perform repository maintenance and state inspection. These tools are standard for the skill's stated purpose of auditing and cleaning Git repositories.
- [INDIRECT_PROMPT_INJECTION]: The skill processes external, potentially untrusted strings from commit messages and pull request metadata which may contain malicious instructions.
- Ingestion points: Repository metadata such as pull request titles and commit logs are ingested via
gh pr listandgit logas defined inSKILL.md. - Boundary markers: The skill includes comprehensive 'Cleanup Rules' and 'Guardrails' sections that explicitly define operational limits and require user approval for all destructive actions.
- Capability inventory: The skill possesses capabilities to delete local branches, drop stashes, remove worktrees, and modify remote pull request states.
- Sanitization: No programmatic sanitization or filtering of the ingested metadata strings is specified; the skill relies on model classification and user review.
Audit Metadata