The Agent Skills Directory

Command Execution: The skill executes a manifest validation command (uv run validate manifest <recipe-path>). This is a legitimate development task used to verify the schema compliance of the generated manifest file. The execution is scoped to the specific recipe directory provided by the user.
File System Access: The skill reads several project files (e.g., pyproject.toml, requirements.txt, app/agent.py) to gather information about the recipe's architecture and language. This read-only access is used exclusively for inferring metadata and does not involve exfiltration of sensitive information.
Hardcoded Placeholders: The skill intentionally uses mandatory placeholders like TODO: Replace with your team name and TODO: Replace with your GitHub user ID. This is a security best practice that prevents the AI from guessing or inferring identity information and ensures that a human developer must provide these credentials before the manifest can be finalized.

generate-manifest