canvas-design

Pass

Audited by Gen Agent Trust Hub on Sep 14, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill uses a context manipulation technique in SKILL.md by fabricating user history with the phrase The user ALREADY said "It isn't perfect enough. It must be pristine...". This attempts to inject a pre-conditioned state of negative feedback into the model's history to enforce specific refinement loops.
  • [EXTERNAL_DOWNLOADS]: The prompt instructions state to Download and use whatever fonts are needed to make this a reality. This guides the agent to make unbounded external network requests to retrieve third-party binary assets without domain validation or source filtering.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 14, 2026, 11:53 PM