hunt-dom

CRITICAL E006: Malicious code pattern detected in skill scripts.

• Malicious code pattern detected (high risk: 1.00). This document contains explicit offensive proof-of-concept code and step-by-step instructions to steal credentials and persistently intercept traffic (postMessage exfil, CSS OOB exfil, and same-origin service-worker backdoor), indicating deliberate malicious/abusive intent.

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow in SKILL.md is an offensive DOM-hunting procedure that requires the LLM to ingest outsider-authored text by explicitly performing attacker-controlled injection tests (e.g., injecting markup, iframe/page content, postMessage payloads, and generating exfil CSS), which can be driven by user-submitted text without selecting a specific trusted item first.