stitch-ui-design-spec-uview
Pass
Audited by Gen Agent Trust Hub on Sep 22, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes user requests (e.g., to 'beautify' or 'refine' a screen) to generate prompts for a downstream agent called Stitch. This multi-step chain creates a potential surface where user-provided content is interpolated into the context of subsequent agent actions, a characteristic of indirect prompt injection vectors. The skill mitigates this by applying a strict contract-based generation process. \n
- Ingestion points: User-provided screen descriptions or refinement requests processed via SKILL.md and references/examples.md. \n
- Boundary markers: The skill structures its output into defined blocks labeled [Context], [Layout], and [Components] to help the downstream agent distinguish between different types of instruction. \n
- Capability inventory: The skill is restricted to generating text and JSON configurations; it contains no subprocess calls, file-write operations, or network interactions. \n
- Sanitization: The skill relies on a predefined library of uView design tokens and component contracts rather than directly executing or passing through arbitrary user-supplied code.
Audit Metadata