wk-worktree-cleanup
Pass
Audited by Gen Agent Trust Hub on Aug 24, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill utilizes standard version control and repository management tools, including
gitand the GitHub CLI (gh). It executes commands such asgit worktree remove,git branch -D, andgit clean -fd. While these commands are destructive by nature, they are used here as intended for repository maintenance. The skill includes explicit instructions to verify that branches are merged (both locally and via GitHub PR status) before proceeding with deletion, and it mandates capturing session context via a 'retro' process to prevent accidental data loss. - [INDIRECT_PROMPT_INJECTION]: The skill processes external data by listing git worktrees and querying GitHub PR titles and statuses.
- Ingestion points: The agent parses output from
git wtl,git status,git log, andgh pr list(specifically file SKILL.md). - Boundary markers: The instructions direct the agent to specifically extract metadata (paths, branch names, PR numbers) rather than interpret or follow content within those strings.
- Capability inventory: All destructive capabilities (file removal and branch deletion) are gated by status checks and conditional logic across SKILL.md and README.md.
- Sanitization: While no explicit regex sanitization is described, the procedural nature of the workflow (checking status codes and boolean merge status) minimizes the risk of the agent following instructions embedded in branch names or PR titles.
Audit Metadata