worktrees

Pass

Audited by Gen Agent Trust Hub on Sep 7, 2026

Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
  • [COMMAND_EXECUTION]: The skill is designed to manage Git worktrees and instructs the agent to execute shell commands such as git worktree add, git worktree remove, and various integration commands like merge or rebase. It incorporates a safety protocol requiring explicit user confirmation for all repository mutations and destructive actions (e.g., git reset --hard, git clean).
  • [INDIRECT_PROMPT_INJECTION]: The skill involves reading state from the local Git repository (via git worktree list) and a local JSON manifest file (.slim/worktrees.json). This creates a surface for indirect prompt injection where a maliciously configured repository or state file could attempt to influence agent actions. Ingestion points: local Git state and .slim/worktrees.json file; Boundary markers: explicit safety guards in SKILL.md requiring user confirmation; Capability inventory: Git CLI command execution; Sanitization: relies on human-in-the-loop verification for all mutation actions.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 7, 2026, 11:24 PM
Security Audit — agent-trust-hub — worktrees