The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes Git commands using subprocess.run with list-based arguments. This approach prevents shell injection vulnerabilities. Operations are confined to the repository scope through explicit path resolution and validation.
[PROMPT_INJECTION]: The skill processes untrusted repository metadata (branch names, commit subjects, file changes), creating an indirect prompt injection surface.
Ingestion points: Data enters the context via git for-each-ref and git show in the maintenance-audit command.
Boundary markers: Documentation requires the agent to review metadata and relation to target before forming a plan.
Capability inventory: Mutating operations include branch merging, deletion, and worktree removal.
Sanitization: Uses list-based subprocess calls for command execution, avoiding shell interpretation.
[SAFE]: The skill implements high-quality security controls, including a mandatory repository lock (agents-worktree.lock) to prevent race conditions during mutations and path validation to ensure worktrees are managed outside the repository root.

git-worktree