composing-grid-canvases
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill contains a significant indirect prompt injection surface because it specifically instructs the agent to use user comments as a source of instructions for fixing or modifying widgets.
- Ingestion points: User feedback and comment threads on canvases retrieved via the
canvas-comments-listandcanvas-comments-retrievetools (SKILL.md). - Boundary markers: None mentioned; the instructions suggest treating comments as a "brief for fixing" without advising on instruction/data separation.
- Capability inventory: The agent has extensive capabilities to modify the environment, including creating components (
canvas-create), retrieving source code (canvas-source-retrieve), and patching layouts (canvas-layout-patch) (SKILL.md). - Sanitization: No sanitization or instruction-ignoring delimiters are specified for the processed comment data.
Audit Metadata