ml-paper-writing
Pass
Audited by Gen Agent Trust Hub on Sep 24, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The
templates/neurips2025/Makefileincludes anupgradetarget that downloads official style files frommedia.neurips.cc. This is a standard and expected operation for maintaining conference template compliance. - [COMMAND_EXECUTION]: The skill provides workflows that use shell tools like
find,grep, andxargsto explore research repositories. These are standard development operations intended to help the agent understand the project context (e.g., finding results and existing citations). - [INDIRECT_PROMPT_INJECTION]: The skill has a large ingestion surface as it processes content from research repositories (READMEs, Python code, results). However, it includes a 'Claim ledger gate' and specific instructions to the agent to verify claims against evidence and mark unsupported statements, mitigating the risk of following adversarial instructions embedded in repo data.
Audit Metadata