canvas-design
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSINDIRECT_PROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: The skill uses a simulated past user interaction within the instructions to steer the agent's behavior towards a specific output quality. Specifically, the text
IMPORTANT: The user ALREADY said "It isn't perfect enough. It must be pristine..."attempts to override current context by fabricating a prior instruction. - [EXTERNAL_DOWNLOADS]: The skill explicitly directs the agent to
Download and use whatever fonts are needed to make this a realitywithout specifying trusted sources or providing validation mechanisms for these external binary resources. - [INDIRECT_PROMPT_INJECTION]: The skill processes user-supplied conceptual threads and niche references to generate visual content. This ingestion of untrusted data, combined with the capability to write files, creates a potential surface for indirect injection where malicious input could attempt to influence the agent's generative logic.
- Ingestion points: User-supplied references and instructions for the design philosophy and canvas creation.
- Boundary markers: None identified in the prompt templates.
- Capability inventory: Writing markdown (.md), PDF (.pdf), and PNG (.png) files; font downloading and usage.
- Sanitization: No specific sanitization or validation of the user-provided conceptual threads is mentioned.
Audit Metadata