animate-watercolor
Pass
Audited by Gen Agent Trust Hub on Nov 1, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill handles untrusted external data such as reference images, captions, and OCR content, which presents a surface for indirect prompt injection.
- Ingestion points: Specified in SKILL.md as reference images, OCR, video captions, and third-party pages.
- Boundary markers: The instructions explicitly direct the agent to treat external content as evidence, not as instructions or authoritative commands.
- Capability inventory: The skill scope includes image editing and repository file access, which are gated by user approval.
- Sanitization: The skill mandates independent validation of commands and destinations for any authorized task.
Audit Metadata