web-design-reviewer

Pass

Audited by Gen Agent Trust Hub on Jul 3, 2026

Risk Level: SAFEPROMPT_INJECTIONCOMMAND_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing untrusted data from external websites.
  • Ingestion points: The agent uses browser_navigate and browser_snapshot to access and read content from external URLs (SKILL.md).
  • Boundary markers: The instructions lack boundary markers or specific warnings to ignore instructions embedded in the external DOM content.
  • Capability inventory: The skill utilizes grep_search, semantic_search, and file write operations to modify the local workspace based on its analysis of the external site (SKILL.md, Step 3.3).
  • Sanitization: No sanitization of the retrieved web content is performed before using it to determine code changes.
  • [COMMAND_EXECUTION]: The skill relies on browser automation tools (e.g., Playwright MCP) to interact with live websites, providing the agent with capabilities to perform network navigation, element interaction, and DOM manipulation.
  • [EXTERNAL_DOWNLOADS]: The documentation provides a configuration example that fetches the Playwright MCP server package using npx from a public registry.
Audit Metadata
Risk Level
SAFE
Analyzed
Jul 3, 2026, 04:01 PM
Security Audit — agent-trust-hub — web-design-reviewer