The Agent Skills Directory

[SAFE]: The skill provides integration guides for well-known LLM services and follows security best practices for production deployments.
[SAFE]: Secrets management consistently uses environment variables (e.g., OPENROUTER_API_KEY, LANGFUSE_SECRET_KEY) instead of hardcoding credentials, preventing accidental exposure of sensitive API keys.
[SAFE]: The calculator tool implementation specifically avoids unsafe evaluation by using the Python ast module to safely parse and evaluate mathematical expressions against a strict whitelist of allowed operators, preventing remote code execution (RCE) vulnerabilities.
[SAFE]: Network operations are directed to well-known and legitimate API endpoints, such as OpenRouter and Langfuse, for their intended purposes of LLM orchestration and observability.
[SAFE]: File system access is restricted to local caching of model responses and temporary processing of document images for OCR, which is standard and expected behavior for the intended functionality.

openrouter-skill