didit-face-match
Pass
Audited by Gen Agent Trust Hub on Oct 13, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [SAFE]: The skill implements a standard integration with the Didit Face Match API. Authentication is handled correctly via environment variables (DIDIT_API_KEY), and all network traffic is directed to legitimate vendor-controlled domains (verification.didit.me, apx.didit.me).
- [INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from the local file system and user-provided metadata, which is then transmitted to an external service.
- Ingestion points: Image files (user_image, ref_image) and configuration strings (vendor_data, metadata) handled by the match_faces function in scripts/match_faces.py.
- Boundary markers: Not utilized; data is passed directly to the API as multipart/form-data.
- Capability inventory: File system read access for images and network egress via the requests library in scripts/match_faces.py.
- Sanitization: The main function in scripts/match_faces.py verifies that the --metadata argument is a valid JSON object.
Audit Metadata