The Agent Skills Directory

[SAFE]: The skill implements advanced network security practices. The provided Python script uses a custom HTTP handler to prevent automatic redirects, ensuring that authentication tokens are not inadvertently sent to unauthorized hosts. It also performs DNS resolution checks to ensure that input and output URLs resolve to global (public) IP addresses, effectively mitigating Server-Side Request Forgery (SSRF) risks against internal networks.\n- [SAFE]: The script enforces strict data boundaries. It implements maximum byte caps for both JSON responses (2MB) and downloaded image assets (30MB), and enforces a maximum pixel count (20 million) before fully decoding images. This protects against resource exhaustion and decompression bomb attacks.\n- [SAFE]: Credential management follows best practices. The skill instructs users to keep the API key server-side and uses environment variables for configuration. The implementation explicitly avoids including the API key in requests to asset or reference hosts.\n- [SAFE]: File operations are handled securely. The script uses atomic writes with temporary files and restricted permissions to manage the job ledger and manifests, preventing race conditions and unauthorized local access.

luma-photon