latex-document

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.80). The skill explicitly fetches BibTeX entries from public DOI/arXiv services (see fetch_bibtex.sh / "Fetch BibTeX from a DOI" in the README/SKILL.md), which ingests untrusted third-party text into .bib files that the agent is expected to read/compile into documents (potentially influencing compilation behavior and allowing injection via LaTeX/BibTeX fields).

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill invokes Mermaid rendering via npx @mermaid-js/mermaid-cli (see https://www.npmjs.com/package/@mermaid-js/mermaid-cli or registry https://registry.npmjs.org/@mermaid-js/mermaid-cli), which causes remote npm package code to be fetched and executed at runtime, meeting the criteria for executing remote code.