The Agent Skills Directory

[COMMAND_EXECUTION]: The skill makes use of common developer tools for its operational logic, including the gh CLI for GitHub interactions and git for worktree management. It also utilizes repository-local scripts (e.g., render_generated_regions.py) and bundled Python Zip Applications (.pyz) for internal processing, which is standard behavior for developer-oriented automation frameworks.
[INDIRECT_PROMPT_INJECTION]: As the skill is designed to ingest and route external data such as pull requests and issues, it has a potential exposure to indirect prompt injection. However, it incorporates a proactive defense in references/coherence-check.md that explicitly instructs the agent to ignore imperative commands found within external content and to prioritize the user's direct request.
[SAFE]: The skill architecture adheres to security best practices for AI agents, including detailed role-based agent resolution (documented in references/agent-resolution.md), capability-based tool discovery (in references/optional-plugins.md), and explicit handling of harness portability concerns. All external URL references are to well-known documentation or service providers, and the system includes built-in safeguards against homoglyph-based status spoofing.

cheese