icp-builder

Warn

Audited by Snyk on Mar 5, 2026

Risk Level: MEDIUM
Full Analysis

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

  • Third-party content exposure detected (high risk: 1.00). The skill's Phase D "Parallel Research" step explicitly instructs launched research agents to "search the web" (fallback to WebSearch) and to gather evidence from public sources including subreddits, Discord servers, forums and other online content, which the agent will read, interpret, score, and use to drive decisions—exposing it to untrusted, user-generated third‑party content.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 5, 2026, 05:19 AM