The Agent Skills Directory

[DATA_EXFILTRATION]: The skill includes functionality to automatically discover and copy sensitive environment files (such as .env, .env.local, and .env.test) from the repository root into newly created worktree directories. This increases the local exposure surface of secrets.
Evidence: In scripts/worktree-manager.sh, the copy_env_files function iterates through all files matching .env* and duplicates them to the target worktree path.
[COMMAND_EXECUTION]: The script executes multiple shell-based Git operations using variables for branch names and paths.
Evidence: The create_worktree, switch_worktree, and cleanup_worktrees functions in scripts/worktree-manager.sh interpolate variables into commands like git worktree add, git checkout, and git worktree remove.
[PROMPT_INJECTION]: The skill has an indirect prompt injection surface because it ingests untrusted branch names and uses them in shell commands without sanitization beyond basic quoting.
Ingestion points: branch_name and from_branch arguments in scripts/worktree-manager.sh.
Boundary markers: Absent.
Capability inventory: git worktree add, git checkout, git pull, and git worktree remove in scripts/worktree-manager.sh.
Sanitization: Absent; the script relies only on double-quoting variables which does not prevent all forms of argument injection.

git-worktree