PAIUpgrade
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: 0.90). The skill explicitly fetches and ingests public third‑party content (e.g., Tools/Anthropic.ts pulls Anthropic blogs, GitHub repos, changelogs and docs; YouTube channels are checked via yt-dlp and VideoTranscript; FindSources searches public YouTube, blogs, GitHub, Discord/Reddit) and the agent is expected to read and analyze that content as part of its workflows, which exposes it to untrusted/user‑generated content and potential indirect prompt injection.
Audit Metadata