canvas-design

Pass

Audited by Gen Agent Trust Hub on Jun 13, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The skill uses behavioral steering in the "FINAL STEP" section by referencing a pre-existing user feedback loop ("The user ALREADY said...") to ensure high craftsmanship. This is a stylistic refinement technique rather than a safety bypass. The skill also processes user input foundations without specific boundary markers or sanitization, representing a minor indirect prompt injection surface.
  • Ingestion points: User-provided design foundations (SKILL.md).
  • Boundary markers: Absent.
  • Capability inventory: File writing and conversion (.pdf, .png, .md).
  • Sanitization: Absent.
  • [EXTERNAL_DOWNLOADS]: The instructions suggest the agent "Download and use whatever fonts are needed." This is a general instruction for creative asset acquisition and does not point to specific untrusted or malicious sources.
  • [COMMAND_EXECUTION]: The skill references searching the local ./canvas-fonts directory for assets, which is a routine and safe file system operation in the context of design work.
Audit Metadata
Risk Level
SAFE
Analyzed
Jun 13, 2026, 06:23 AM