web-design-guidelines

Warn

Audited by Gen Agent Trust Hub on Aug 16, 2026

Risk Level: MEDIUMPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill metadata identifies the author as 'vercel', which contradicts the actual author account 'eva813'. This form of metadata poisoning can mislead users about the skill's provenance and security profile.\n- [EXTERNAL_DOWNLOADS]: Fetches external guidelines from 'https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md'. Retrieving content that dictates agent behavior at runtime from a remote source introduces a surface for instruction manipulation.\n- [INDIRECT_PROMPT_INJECTION]: The skill processes both external guidelines and user-provided code files without explicit boundary markers or sanitization logic.\n
  • Ingestion points: External guidelines file and user-specified UI code files referenced in SKILL.md.\n
  • Boundary markers: Absent; there are no delimiters to separate the rules/data from instructions.\n
  • Capability inventory: The agent uses network capabilities (WebFetch) and has read access to the local file system.\n
  • Sanitization: No specific validation or filtering of the content retrieved from the external URL is described.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Aug 16, 2026, 12:30 AM
Security Audit — agent-trust-hub — web-design-guidelines