canvas-design

Pass

Audited by Gen Agent Trust Hub on Feb 27, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill uses persona-reinforcement techniques in the 'FINAL STEP' section, simulating a user request ('The user ALREADY said...') to ensure the agent prioritizes 'museum quality' craftsmanship. This is a steering mechanism for output quality rather than an attempt to bypass safety guidelines.
  • [EXTERNAL_DOWNLOADS]: The skill mentions downloading external fonts if needed. This is a functional requirement for graphic design tasks, and no suspicious or specific third-party URLs are hardcoded.
  • [INDIRECT_PROMPT_INJECTION]:
  • Ingestion points: The agent is instructed to 'deduce the subtle conceptual thread' from niche references embedded in user prompts (SKILL.md).
  • Boundary markers: Absent. The skill does not define explicit delimiters to separate user data from its internal reasoning instructions.
  • Capability inventory: The agent possesses the capability to generate and export content as .md, .pdf, and .png files.
  • Sanitization: Absent. The agent relies on its internal safety filters while interpreting the user's conceptual references for visual expression.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 27, 2026, 07:59 AM