The Agent Skills Directory

[PROMPT_INJECTION]: The skill implements strong defensive measures against prompt injection. Both the master generation templates and the runtime retrieval instructions include explicit directives to treat external data as non-executable material. All pre-built SKILL.md files include mandatory sections that define non-negotiable behavior rules for the agent.
[DATA_EXFILTRATION]: No exfiltration of sensitive data was found. Network operations are limited to fetching religious texts from the project's official backend (fojin.app), which is consistent with the skill's purpose. The code does not attempt to access sensitive local files or environment variables.
[SAFE]: No obfuscated content or malicious hidden instructions were detected. The codebase includes specialized logic in master_builder.py, rag_query.py, and skill_writer.py to proactively sanitize Unicode control characters (such as zero-width spaces or bidi markers) and recursively remove forged data markers from processed content.
[REMOTE_CODE_EXECUTION]: The skill generates static markdown and metadata files rather than executable scripts. All third-party dependencies are well-known, legitimate libraries pinned to specific versions. The project follows supply chain security best practices, including SHA-pinning all GitHub Actions and employing OIDC Trusted Publishing for its npm distribution. The Python and Node.js components include rigorous charset validation to prevent path traversal during file operations.

create-master