canvas-design
Pass
Audited by Gen Agent Trust Hub on Oct 6, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The SKILL.md file contains a 'user-already-said' override pattern in its refinement instructions, which attempts to manipulate the agent's conversation state to force a specific high-quality output style.
- [EXTERNAL_DOWNLOADS]: The skill instructs the agent to 'Download and use whatever fonts are needed', which facilitates network requests for external binary data. The skill includes license documentation pointing to well-known repositories from trusted sources like Google, Vercel, and IBM.
- [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied conceptual inputs to generate visual philosophies and final designs, creating a potential surface for indirect injection. 1. Ingestion points: User-provided instructions and conceptual threads in SKILL.md. 2. Boundary markers: Absent. 3. Capability inventory: File writing (.md, .pdf, .png) and external asset downloading. 4. Sanitization: None identified.
Audit Metadata