The Agent Skills Directory

[COMMAND_EXECUTION]: The skill is designed to interact with the local operating system by executing Git CLI commands and navigating directories to manage parallel work environments.
Evidence: Execution of commands such as git worktree add, git branch -d, git push, and cd documented throughout SKILL.md.
Safety context: The skill enforces a safety policy requiring the agent to obtain explicit user confirmation before executing any deletion commands (git worktree remove, git branch -d, etc.).
[INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied task descriptions to generate branch names and directory paths, which are then used in shell commands.
Ingestion points: User-provided task names are ingested to form directory names like ../.wt-<task> and branch names like feature/wx-<task>.
Boundary markers: The skill includes a mandatory instruction that "一切删除动作执行前必须先向用户确认" (all deletion actions must first be confirmed by the user).
Capability inventory: The skill utilizes subprocess execution of Git commands and file system navigation via cd in SKILL.md.
Sanitization: No explicit shell-escaping or validation logic for the task names is defined in the instructions, relying on the user confirmation step for safety.

worktree