ml-paper-writing
Pass
Audited by Gen Agent Trust Hub on Apr 11, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONPROMPT_INJECTIONDATA_EXFILTRATION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Fetches conference style configuration files from the official conference domain (
media.neurips.cc) and GitHub repositories of recognized academic organizations (ACL, COLM). - [EXTERNAL_DOWNLOADS]: Recommends integrating the Exa MCP server from
exa.aifor academic searching. - [COMMAND_EXECUTION]: Uses standard shell commands (
find,grep,ls) to analyze research repository structures and compile documents using LaTeX toolchains (pdflatex,latexmk). - [DATA_EXFILTRATION]: Interacts with Semantic Scholar, CrossRef, and arXiv APIs to verify citation data and retrieve BibTeX entries, which is consistent with the skill's academic objectives and involves no sensitive local data.
- [PROMPT_INJECTION]: The skill processes contents of user research repositories (READMEs, documentation) to draft papers, which presents an indirect prompt injection surface if processed files contain malicious instructions. The risk is minimized by the skill's focus on evidence-based drafting and citation verification.
Audit Metadata