canvas-design
Pass
Audited by Gen Agent Trust Hub on Sep 25, 2026
Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses a behavioral override technique in the 'FINAL STEP' section, instructing the agent to assume 'The user ALREADY said "It isn't perfect enough. It must be pristine..."'. This attempts to inject a specific user state into the agent's context to force behavior that ignores the actual current user session.- [COMMAND_EXECUTION]: The instructions direct the agent to perform file system and network operations, including searching specific local paths ('./canvas-fonts') and fetching external resources ('Download and use whatever fonts are needed').- [INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and interpret 'subtle input or instructions by the user' to generate multi-format outputs (.md, .pdf, .png).
- Ingestion points: User-supplied creative briefs or instructions.
- Boundary markers: Absent; the skill encourages 'full creative freedom' and interpreting 'subtle references' without delimiters.
- Capability inventory: File system read/write access and network download capabilities for assets.
- Sanitization: None specified for user-provided strings before they are incorporated into the 'visual philosophy' or rendered documents.
Audit Metadata