reviewing-interface-quality

Pass

Audited by Gen Agent Trust Hub on Aug 3, 2026

Risk Level: SAFEPROMPT_INJECTION
Full Analysis
  • [PROMPT_INJECTION]: The skill ingests content from external or local web pages via screenshots, DOM extraction, and computed style analysis, creating a surface for indirect prompt injection where malicious instructions embedded in a UI could influence the agent's audit report.
  • Ingestion points: Web content is ingested via Playwright in SKILL.md (e.g., http://localhost:5173).
  • Boundary markers: No explicit delimiters or instructions are provided to the agent to distinguish between its own instructions and text found within the audited UI.
  • Capability inventory: The skill writes screenshots to the local file system (/tmp/) and executes shell commands for code analysis (grep) and accessibility auditing (npx @axe-core/cli).
  • Sanitization: No sanitization is performed on the computed styles or DOM elements extracted from the web page context.
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 3, 2026, 10:47 AM
Security Audit — agent-trust-hub — reviewing-interface-quality