github-deep-research
Warn
Audited by Snyk on Apr 28, 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). Yes — the SKILL.md research workflow (Rounds 2–3) explicitly instructs using web_search and web_fetch plus the GitHub API to ingest public, user-generated sources (repo READMEs, issues/PRs, Reddit/HN/social media) which the agent must read and use to drive analysis and actions, exposing it to untrusted third‑party content that could contain injected instructions.
Issues (1)
W011
MEDIUMThird-party content exposure detected (indirect prompt injection risk).
Audit Metadata