The Agent Skills Directory

[SAFE]: Strict configuration management. The skill implements a robust security boundary by requiring an explicit agent_config.toml file where all providers, endpoints, and models must be manually whitelisted. Any attempt to invoke a model or endpoint not present in this configuration is rejected locally before any network request is initiated.\n- [SAFE]: Secure credential management. API keys are managed through environment variables rather than being hardcoded. The skill includes a dedicated redaction mechanism in its service layer that scans and removes API keys from error messages and logs, preventing accidental exposure of sensitive credentials.\n- [SAFE]: Input validation and capability enforcement. The skill performs deep validation of all request parameters—such as image dimensions, quality settings, and output formats—against a model-specific capability matrix. Parameters not explicitly supported by the selected model's policy are rejected, ensuring the agent adheres to least-privilege principles.\n- [SAFE]: Verified dependencies and execution. The skill uses industry-standard Python libraries (httpx, pillow) and runs its logic via the uv tool on locally bundled scripts. All network communication is restricted to the official, well-known API endpoints of the configured providers.

image-generation