The Agent Skills Directory

[COMMAND_EXECUTION]: The skill automatically executes environment setup and testing commands including npm install, pip install, cargo build, poetry install, and various test runners (pytest, cargo test, etc.) after initializing a worktree. This is standard behavior for a development-focused skill but involves executing code determined by the contents of the repository.
[EXTERNAL_DOWNLOADS]: The use of package managers such as npm, pip, poetry, and go involves downloading third-party dependencies from public registries.
[PROMPT_INJECTION]: The skill parses CLAUDE.md to find worktree directory preferences. This introduces an indirect prompt injection surface where repository content can influence the agent's behavior.
Ingestion points: Reads configuration from CLAUDE.md using grep (SKILL.md).
Boundary markers: None identified; the tool directly searches for pattern matches in the file.
Capability inventory: Performs git operations, file system modifications (creating directories, editing .gitignore), and executes arbitrary build/test commands (SKILL.md).
Sanitization: None identified for the input read from the project file.

using-git-worktrees