influencer-marketing
Pass
Audited by Gen Agent Trust Hub on Oct 5, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted external data (creator bios, captions, transcripts), creating an indirect prompt injection surface.
- Ingestion points: External content retrieved via social-fetch (SKILL.md).
- Boundary markers: The skill explicitly warns the agent to ignore instructions embedded in fetched content.
- Capability inventory: Retrieval tools like social-fetch and watch-video are used.
- Sanitization: The skill employs instructional guardrails to prevent the agent from following external commands.
Audit Metadata