arxiv-doc-builder

Warn

Audited by Snyk on May 5, 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). This skill automatically fetches public arXiv content (see SKILL.md and fetch_paper.py which curl arxiv.org URLs) and explicitly instructs using Claude's Read/vision tools to read converted pages (convert_pdf_with_vision.py), so untrusted, user-submitted paper content will be ingested and interpreted as part of the agent's workflow and could therefore influence subsequent tool use and decisions.

Issues (1)

W011
MEDIUM

Third-party content exposure detected (indirect prompt injection risk).

Audit Metadata
Risk Level
MEDIUM
Analyzed
May 5, 2026, 04:23 PM
Issues
1