deep-research

Warn

Audited by Snyk on Feb 21, 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 SKILL.md activation and "Collect sources" steps require fetching and ingesting external links and public web sources (official docs, blogs, changelogs and even Reddit/Discord per references/source-quality.md), which the agent must read and synthesize into recommendations, exposing it to untrusted third-party content that could carry indirect prompt injection.
Audit Metadata
Risk Level
MEDIUM
Analyzed
Feb 21, 2026, 11:21 AM