canvas-design

Pass

Audited by Gen Agent Trust Hub on Aug 30, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied creative direction as the primary input for generating visual designs and aesthetic manifestos.
  • Ingestion points: User instructions are ingested via the agent's standard interaction loop and used to define the 'Visual Philosophy' (SKILL.md).
  • Boundary markers: The skill lacks explicit delimiters or instructions to ignore potential system commands embedded within the user's design requests.
  • Capability inventory: The skill can generate and save files (.md, .pdf, .png) and is instructed to search for or download external assets (fonts).
  • Sanitization: No input sanitization or validation mechanisms are mentioned for the incoming user data.
  • [EXTERNAL_DOWNLOADS]: The skill provides explicit instructions to fetch external assets at runtime to support its design tasks.
  • Evidence: SKILL.md contains the instruction: 'Download and use whatever fonts are needed to make this a reality.'
  • Context: While no specific malicious URLs are provided, this encourages the agent to interact with external servers for binary font files.
  • Trusted references: The skill includes numerous font license files (OFL) in the canvas-fonts/ directory that reference official repositories of well-known organizations, such as Vercel (Geist) and IBM (Plex).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 30, 2026, 02:56 PM
Security Audit — agent-trust-hub — canvas-design