ml-paper-writing
Pass
Audited by Gen Agent Trust Hub on May 19, 2026
Risk Level: SAFECOMMAND_EXECUTIONEXTERNAL_DOWNLOADSREMOTE_CODE_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill incorporates instructions and automation for standard build tools used in academic document preparation.
- File
templates/README.mdand theneurips2025Makefile include commands for compiling LaTeX documents usingpdflatex,bibtex,latexmk, andpdfcrop. - Instructions in
SKILL.mdguide the agent to explore local research repositories using standard shell utilities likels,find, andgrepto identify experimental results and relevant citations. - [EXTERNAL_DOWNLOADS]: The skill provides mechanisms for obtaining official conference style files and documentation from reputable sources.
- The
templates/neurips2025/Makefileincludes an upgrade target that usescurlto fetch style packages directly frommedia.neurips.cc. - Documentation across the templates refers to official repositories on GitHub (e.g.,
github.com/acl-org) and cloud-based platforms like Overleaf for maintaining current versions of style files. - [REMOTE_CODE_EXECUTION]: The skill includes Python scripts and platform-specific commands to automate research-related data fetching.
SKILL.mdandreferences/citation-workflow.mdcontain Python code snippets using therequests,semanticscholar, andarxivlibraries to fetch paper metadata and BibTeX entries from established academic APIs atdoi.organdsemanticscholar.org.SKILL.mdalso includes an instruction to add an academic search capability using the commandnpx -y mcp-remote "https://mcp.exa.ai/mcp".- [SAFE]: The skill's functionality is transparently documented and adheres to the principles of legitimate research productivity tools. No malicious patterns such as credential harvesting, data exfiltration to untrusted domains, or attempts to override agent safety constraints were identified.
Audit Metadata