The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses subprocess.run to interact with git and the GitHub CLI (gh). This is fundamental to its purpose as a repository management tool. Usage is constrained to a list-based argument format, which mitigates command injection risks.- [INDIRECT_PROMPT_INJECTION]: The tool ingests pull request metadata from GitHub. While the skill identifies this as untrusted data and instructs the agent to treat it as source material rather than instructions, there is an inherent attack surface. The risk is minimized by the tool's focus on SHAs and filesystem states rather than natural language interpretation.- [DATA_EXFILTRATION]: The script analyzes local files, including sensitive ones like .env, to ensure they are not accidentally deleted if unique to a worktree. It does not transmit these secrets externally, only performing local comparisons and branch operations via authenticated git and GitHub tools.

git-cleanup