fal-media
Pass
Audited by Gen Agent Trust Hub on Nov 7, 2026
Risk Level: SAFENO_CODEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill retrieves instructions from an external community library using the
read_guidancetool, creating a surface for indirect prompt injection. \n - Ingestion points: Untrusted data enters the agent context through the
read_guidancetool as specified in SKILL.md. \n - Boundary markers: The skill includes explicit instructions for the agent to treat returned guidance as reference material rather than new permissions. \n
- Capability inventory: The skill possesses capabilities for local file uploads and the execution of paid media generation jobs. \n
- Sanitization: The skill mandates that all guidance must be verified against live production contracts and model schemas before implementation. \n- [REMOTE_CODE_EXECUTION]: The skill contains a specific security directive prohibiting the agent from fetching or running external scripts to recreate missing host environments.
Audit Metadata