stitch-design-taste
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted user data, specifically the 'target project's intent' and 'user's vibe description', which are used to generate content for
DESIGN.mdand local wiki entries. - Ingestion points: User-provided project intent and vibe descriptions used to adapt design parameters (SKILL.md, Section 1).
- Boundary markers: Absent. The skill does not instruct the agent to use delimiters or warnings to ignore instructions that might be embedded in the user's design description.
- Capability inventory: The skill utilizes file-writing capabilities to generate a
DESIGN.mdfile and to maintain a local wiki (modifyingllmwiki/wiki/index.md,llmwiki/wiki/log.md, and creating new draft files). - Sanitization: Absent. There are no instructions to sanitize or validate the user-supplied input before interpolation into the generated artifacts.
Audit Metadata