The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes a series of shell commands to manage Git worktree lifecycles, including git worktree add, git rev-parse, and directory operations. It also automates project-specific build and test routines such as npm test, cargo test, and go test, which are essential for its primary function of environment setup.
[EXTERNAL_DOWNLOADS]: To prepare the isolated workspace, the skill triggers automated dependency installation using standard package managers including npm, pip, poetry, and go mod. These operations fetch code from official package registries and represent standard developer workflows.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface (Category 8) by consuming configuration data from CLAUDE.md and project-specific files like package.json or Cargo.toml. This allows external project content to influence the agent's execution path.
Ingestion points: Reads directory preferences from CLAUDE.md and setup requirements from project manifest files.
Boundary markers: None explicitly implemented for internal file content parsing.
Capability inventory: Includes file system modification (.gitignore), Git commits, network operations via package managers, and arbitrary command execution for testing.
Sanitization: Relies on git check-ignore for basic safety but lacks sanitization for instructions parsed from configuration files.

using-git-worktrees