qa-browser-data-collector

Pass

Audited by Gen Agent Trust Hub on Sep 18, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill is designed to autonomously explore and extract data from external, untrusted web applications via Playwright. This creates a surface for indirect prompt injection attacks where a malicious website could embed instructions to override the agent's behavior.
  • Ingestion points: Data is ingested from live web application page structures, forms, accessibility snapshots, and network API responses (SKILL.md).
  • Boundary markers: The instructions lack specific delimiters or "ignore embedded instructions" warnings to isolate the agent from instructions found within target web pages.
  • Capability inventory: The agent utilizes Playwright MCP tools for browser navigation, interaction (clicks, hovers), and network traffic recording (references/playwright-mcp-patterns.md).
  • Sanitization: There is no mention of sanitizing, escaping, or filtering content retrieved from the web before the agent processes it into a UI specification.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 18, 2026, 05:38 PM
Security Audit — agent-trust-hub — qa-browser-data-collector