ml-paper-writing

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

• Third-party content exposure detected (high risk: 0.90). The SKILL.md and references/citation-workflow.md explicitly instruct the agent to search and fetch content from public third-party sources (e.g., Exa MCP at https://mcp.exa.ai/mcp, Semantic Scholar, CrossRef, arXiv) and to read/verify those papers and BibTeX programmatically as part of the citation and drafting workflow, which lets untrusted external content materially influence subsequent drafting and tool use.