anthropic-canvas-design
Pass
Audited by Gen Agent Trust Hub on Mar 7, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Behavioral manipulation through fabricated history. The SKILL.md file includes a 'FINAL STEP' section that falsely claims the user has already provided negative feedback ('The user ALREADY said...') to force the agent into a specific refinement loop.
- [PROMPT_INJECTION]: Use of high-priority override markers. The prompt frequently uses 'CRITICAL' and 'IMPORTANT' to prioritize its specific design constraints over standard agent instructions.
Audit Metadata