nemo-rl-auto-research
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFECOMMAND_EXECUTION
Full Analysis
- [COMMAND_EXECUTION]: The skill instructs the agent to execute NeMo-RL training and evaluation jobs using
uv run,launch-nemo-rl, and native Slurm launchers. This is the core functionality of the research agent and is governed by explicit instructions to seek user confirmation before initiating any compute-intensive operations or making significant changes to the repository's git state. - [PROMPT_INJECTION]: The skill workflow involves reading external research recipes and source code, which constitutes an indirect prompt injection surface. This is a standard requirement for an autonomous research agent, and the instructions mitigate potential risks by requiring human-in-the-loop verification of campaign plans and specific commands before they are executed.
Audit Metadata