The Agent Skills Directory

[SAFE]: The skill enforces strict security and operational guidelines for managing sensitive data.
It specifies server-side storage for API keys and recommends restricting their scope.
It mandates the deletion of remote File API resources and temporary local files after use to minimize the data footprint.
[SAFE]: Network operations are restricted to trusted infrastructure.
The implementation uses official Google API endpoints (generativelanguage.googleapis.com and aiplatform.googleapis.com).
The reference script includes a PinnedHTTPSConnection class to verify endpoint identity and resolution before transmission.
[SAFE]: Robust validation logic is implemented for generated artifacts.
The skill uses ffprobe and ffmpeg to programmatically verify video parameters such as FPS, resolution, and duration.
Cryptographic hashing (SHA-256) is used to maintain a ledger of source and output media, ensuring integrity and provenance.
[SAFE]: A fail-closed transaction model prevents unauthorized or redundant spending.
The workflow requires a unique attempt UUID and a cryptographic hash of the approved plan.
The script uses an atomic attempt log to prevent concurrent or replayed executions of paid API calls in the event of connection failures.

google-gemini-omni-video