The Agent Skills Directory

[SAFE]: The skill defines structural rules for git worktree isolation and does not contain any malicious patterns or unauthorized data access.
[COMMAND_EXECUTION]: The skill instructs the agent to use standard git commands (git checkout, git rev-parse, git add) for local repository management within a controlled worktree environment.
[PROMPT_INJECTION]: Behavioral instructions such as 'STOP and report the error' are used as safety guardrails to ensure the agent is operating in the correct environment, rather than attempting to override core safety filters or system instructions.

worktree-agent