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 fetches an instruction file from an external URL to determine its behavior at runtime.
  • Evidence: The skill uses WebFetch to retrieve https://raw.githubusercontent.com/vercel-labs/web-interface-guidelines/main/command.md.
  • Trust Context: The repository vercel-labs/web-interface-guidelines is part of the vercel-labs organization, which is a Trusted External Source. Per the [TRUST-SCOPE-RULE], this finding is downgraded to LOW.
  • PROMPT_INJECTION (LOW): The skill exhibits a surface for indirect prompt injection by processing untrusted data (user files) alongside remote instructions.
  • Ingestion points: The skill reads external content from a GitHub URL and user-provided files/patterns (SKILL.md).
  • Boundary markers: Absent; there are no explicit delimiters or instructions to ignore embedded commands within the files being reviewed.
  • Capability inventory: The skill uses WebFetch for network reads and has the ability to read local files via the agent's filesystem tools.
  • Sanitization: Absent; the fetched content and user files are used directly to influence the agent's output logic.
  • DATA_EXFILTRATION (SAFE): The skill reads local files for the purpose of a UI review. While it performs a network request to fetch guidelines, it does not send the contents of the local files to an external endpoint; the analysis results are returned to the user.
Audit Metadata
Risk Level
SAFE
Analyzed
Feb 17, 2026, 06:10 PM