problem-solving

Warn

Audited by Snyk on Feb 27, 2026

Risk Level: MEDIUM
Full Analysis

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

  • Third-party content exposure detected (high risk: 0.90). The skill's Collision-Zone Thinking instructions (references/collision-zone-thinking.md, "Deep Research" step) explicitly require studying open web sources — e.g., "read industry blogs/news" and "join Reddit communities" — which directs the agent/workflow to ingest and act on untrusted, user-generated third-party content that can influence decisions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 27, 2026, 10:26 PM