The Agent Skills Directory

[COMMAND_EXECUTION]: The skill generates scripts that utilize various system utilities including gh (GitHub CLI) for secret management, python3 for environment file manipulation, and platform-specific commands like open, xdg-open, or wslview to open documentation URLs in the user's browser.
[INDIRECT_PROMPT_INJECTION]: The skill acts as a template generator that consumes repository artifacts and user-provided instructions to construct the wizard stages, creating a potential surface for indirect prompt injection.
Ingestion points: The agent reads local repository files such as README.md, .env.example, and workflow configurations to determine necessary manual setup steps.
Boundary markers: The skill relies on the agent's reasoning to correctly interpret instructions without explicit delimiter-based enforcement against embedded malicious instructions in the source documentation.
Capability inventory: The generated scripts have the capability to write to local files (.env), set repository-level secrets on GitHub, and open external URLs.
Sanitization: The included envfile.py script validates environment variable keys and encodes values to ensure they are stored safely in a literal format, preventing syntax-based injection into environment files.

wizard