The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes local Git and shell commands to facilitate development workflow. * Evidence: The script uses git rev-parse, mkdir, and git worktree add to create isolated project environments. * Context: All operations are conducted locally within the user's home directory and existing Git repositories.
[PROMPT_INJECTION]: The skill identifies a potential indirect injection surface from user-provided feature names but includes mitigation steps. * Ingestion points: User-supplied <feature> name via prompt. * Boundary markers: Not explicitly used in shell snippets. * Capability inventory: Command execution via git and mkdir. * Sanitization: The skill documentation explicitly instructs the agent to 'normalize to a short dash-cased feature string' before creating the worktree, which prevents shell metacharacter injection.

using-git-worktrees