eval-harness-updater

Warn

Audited by Snyk on Mar 3, 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 Research Gate explicitly requires using Exa for web_search/get_code_context and even a direct WebFetch to arxiv.org (see "Research Gate (Exa + arXiv — BOTH MANDATORY)" and the WebFetch example), so the agent will fetch and interpret public web/arXiv content that can influence harness changes, exposing it to untrusted third-party instructions.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Mar 3, 2026, 02:59 AM