The Agent Skills Directory

[COMMAND_EXECUTION] (SAFE): The skill executes multiple shell commands including git worktree, npm install, cargo build, and pip install. While these commands execute code, they are standard for the primary purpose of workspace initialization.
[EXTERNAL_DOWNLOADS] (SAFE): Invokes package managers (npm, pip, poetry, go mod) which download external dependencies. This is expected behavior for setting up a development environment.
[INDIRECT_PROMPT_INJECTION] (LOW):
Ingestion points: The skill reads user preferences from a local file named CLAUDE.md using grep to determine directory locations.
Boundary markers: No specific boundary markers are used when reading from CLAUDE.md.
Capability inventory: The skill can execute arbitrary code via package manager installation scripts and test runners (npm test, pytest, etc.).
Sanitization: The skill employs a case statement to validate the extracted LOCATION variable, ensuring it matches expected patterns (.worktrees, worktrees, or a specific global path) before using it in path construction.

using-git-worktrees