canvas-design
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- [PROMPT_INJECTION]: The skill employs a 'simulated history' technique by instructing the agent to act as if the user 'ALREADY said' the work isn't perfect enough, which is a method used to force a specific refinement behavior regardless of actual user input.
- [PROMPT_INJECTION]: The instructions command the agent to hide its AI provenance by ensuring the output looks 'human-crafted' and 'not AI-generated,' effectively using role-play to bypass transparency or safety markers associated with AI-generated content.
- [EXTERNAL_DOWNLOADS]: The agent is encouraged to 'Download and use whatever fonts are needed,' which introduces a risk of fetching assets from unverified or untrusted external sources at runtime.
- [PROMPT_INJECTION]: The skill uses aggressive, imperative language ('non-negotiable,' 'MUST stress,' 'CRITICAL GUIDELINES') to override the agent's standard creative boundaries and force adherence to a specific 'master craftsman' persona.
Audit Metadata