The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the gh (GitHub CLI) and git tools to perform operations such as editing repository descriptions, creating labels, and managing repository settings. The skill explicitly mandates preflight authentication checks and requires a manual 'OK' from the user before running any commands that modify the repository state.
[INDIRECT_PROMPT_INJECTION]: The skill processes untrusted repository content (such as README.md and manifest files like package.json or pyproject.toml) to generate one-line descriptions and topic tags. This ingestion of external data constitutes an attack surface where maliciously crafted repository content could attempt to influence the agent's output.
Ingestion points: Reads README.md and project manifests (package.json, pyproject.toml, Cargo.toml, go.mod, etc.) in modes/about.md and modes/setup.md.
Boundary markers: None explicitly defined for the data ingestion process.
Capability inventory: The skill uses gh repo edit, gh label create/edit, and the Write tool to modify the repository and local file system.
Sanitization: No specific sanitization techniques are detailed, but the mandatory user preview and approval for all proposed changes serves as a primary mitigation.
[EXTERNAL_DOWNLOADS]: The skill fetches license templates and user metadata from official GitHub API endpoints (e.g., gh api /licenses/<key>) to populate repository files. These operations target a well-known service for their intended purpose.

repokit