worktree-hygiene
Pass
Audited by Gen Agent Trust Hub on Oct 17, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is designed to manage Git worktrees safely, preventing data loss and process clobbering through strictly defined gates and validation steps.\n- [COMMAND_EXECUTION]: The skill includes a helper script (
scripts/retire-branch.mts) that automates branch retirement using Git commands,lsof, andps. The script uses security best practices by executing commands via argument arrays rather than shell strings and employing the--end-of-optionsdelimiter for Git commands to prevent argument injection from malicious branch names.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes external data including Git branch names, file paths, and process lists.\n - Ingestion points: Data enters the script via Git metadata and system process utilities from the local repository environment.\n
- Boundary markers: The script uses
--end-of-optionsto explicitly delimit branch names and paths in Git command invocations, preventing malicious names from being interpreted as flags.\n - Capability inventory: The script has the capability to fetch from Git remotes, remove local worktree directories, and delete local Git branch references.\n
- Sanitization: All external inputs used in commands are handled through array-based spawning (child_process.spawnSync) and parameter delimitation, neutralizing potential shell or argument injection vectors.
Audit Metadata