parallel-web-search

Warn

Audited by Snyk on Feb 23, 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.md explicitly instructs calling Parallel AI's web search API ("Search Request" / "Response") to fetch open-web results (titles, URLs, excerpts) — including content from public sites (see source_policy examples mentioning reddit/twitter) — which the agent is expected to read and act on (mode: "agentic"), so untrusted third‑party content can materially influence behavior.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 23, 2026, 12:01 PM