canvas-design
Pass
Audited by Gen Agent Trust Hub on Mar 6, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill employs a 'pre-emptive user simulation' technique, instructing the agent to act as if the user has already requested a 'masterpiece' version of the art. This simulates conversation history to override standard agent behavior and force a refinement loop.
- [PROMPT_INJECTION]: The skill processes user-defined conceptual themes to influence the design process, creating an indirect prompt injection surface.
- Ingestion points: User-provided 'subtle conceptual threads' ingested during the 'deducing the subtle reference' step.
- Boundary markers: Absent; the user input is interpolated into the creative process without specific delimiters or warnings to ignore embedded instructions.
- Capability inventory: Limited to the generation of static document and image files including .md, .pdf, and .png.
- Sanitization: No validation or escaping of user input is performed before it is used to influence the design philosophy.
Audit Metadata