The Agent Skills Directory

[SAFE]: The skill follows security best practices by explicitly instructing the agent never to hardcode API keys or secrets in the generated code or documentation, recommending the use of environment variables instead.
[COMMAND_EXECUTION]: The skill leverages Git and the GitHub CLI (gh) to automate repository initialization and publication. This functionality is mitigated by mandatory instructions requiring the agent to obtain explicit user confirmation for repository names, visibility settings, and execution timing.
[INDIRECT_PROMPT_INJECTION]: The skill processes user-defined themes to generate project code, creating a surface for indirect prompt injection.
Ingestion points: User-provided theme in Step 1 (SKILL.md).
Boundary markers: Specific instructions to treat the theme as data and disregard any embedded commands.
Capability inventory: File system writes (server.py, README.md), Git commands, and GitHub CLI execution.
Sanitization: Relies on the agent to perform semantic filtering and observe suspicious input patterns.
[CREDENTIALS_UNSAFE]: The generated .gitignore template proactively includes patterns for .env and other environment files to prevent accidental credential leakage in version control.

remote-mcp