Social Video
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes data from various local project files, such as brand-identity.md, ICP.md, and messaging.md, to generate media briefs and drafts. This design presents an indirect prompt injection surface where malicious instructions embedded in these documents could be influenced by the agent. * Ingestion points: Brand documentation and campaign briefs located in the workspace. * Boundary markers: Not present; instructions do not specify delimiters or warnings for processed content. * Capability inventory: File reading, workspace writing, and media generation/upload via MuAPI. * Sanitization: Not present; the skill lacks validation or filtering for the content of ingested files.
Audit Metadata