research-after-failure

Warn

Audited by Snyk on Feb 15, 2026

Risk Level: MEDIUM
Full Analysis

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

  • Third-party content exposure detected (high risk: 0.80). The skill's "Step 4: Research Online" explicitly directs the agent to search and read public web sources (e.g., exact error text searches, Stack Overflow, GitHub Issues and other web results), which are untrusted, user-generated third‑party content that the agent is expected to ingest and act on.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 15, 2026, 09:44 PM