canvas-design
Pass
Audited by Gen Agent Trust Hub on Apr 12, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill includes instructions that simulate specific user feedback ('The user ALREADY said...') intended to override the agent's deliberation and force a 'masterpiece' quality level.
- [EXTERNAL_DOWNLOADS]: The skill directs the agent to download external fonts from the internet, introducing unverified third-party assets and potential supply chain risks.
- [PROMPT_INJECTION]: The skill processes untrusted user data ('subtle references') without sanitization or boundary markers while maintaining capabilities for file writing and network downloads. Evidence Chain: 1. Ingestion point: User input foundation for art creation. 2. Boundary markers: Absent. 3. Capability inventory: File system writes and network download capabilities. 4. Sanitization: Absent.
Audit Metadata