cleanup-worktrees

Pass

Audited by Gen Agent Trust Hub on Jun 16, 2026

Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill uses shell commands to interact with git and the GitHub CLI. It implements safety rules such as requiring user confirmation and skipping worktrees with uncommitted changes.\n- [PROMPT_INJECTION]: The skill processes untrusted output from git commands (branch names and paths) and interpolates them into subsequent shell commands. This creates a surface for indirect prompt injection (Category 8) if the agent handles a malicious repository. 1. Ingestion points: Output of git worktree list --porcelain (SKILL.md). 2. Boundary markers: Absent. 3. Capability inventory: git worktree remove, git branch -d. 4. Sanitization: None; the skill relies on variable interpolation with standard shell quoting.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 16, 2026, 06:20 PM
Security Audit — agent-trust-hub — cleanup-worktrees