web-design-guidelines
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- EXTERNAL_DOWNLOADS (LOW): The skill is configured to fetch a remote markdown file from GitHub to define its logic and output format at runtime.
- Evidence: Fetching from
https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md. - Trust Status: The organization
vercel-labsis a trusted source. Per [TRUST-SCOPE-RULE], the severity of the remote fetch is downgraded to LOW. - PROMPT_INJECTION (LOW): There is a potential for indirect prompt injection as the agent ingests both remote instructions and untrusted local UI code files.
- Ingestion points: Fetched
command.mdfile and user-specified local files. - Boundary markers: Not specified in the skill description; the agent relies on its own ability to distinguish between instructions and data.
- Capability inventory: Network read (WebFetch) and local file read access.
- Sanitization: No explicit sanitization or escaping of the fetched content or user files is mentioned.
Audit Metadata