nemo-gym-reward-profiling
Pass
Audited by Gen Agent Trust Hub on Jul 22, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides documentation and shell command templates for Nemo Gym reward profiling. The content is instructional and does not contain any hidden or malicious logic.
- [COMMAND_EXECUTION]: The skill includes bash script templates in
references/quick-start.mdfor running Nemo Gym commands likegym env startandgym eval run. These commands are standard for the tool's usage and include placeholders for user-provided credentials, ensuring that secrets are not hardcoded or automatically exposed. - [PROMPT_INJECTION]: The skill describes a workflow that involves reading and processing external data from JSONL files (
rollouts.jsonland*_materialized_inputs.jsonl). While this presents a theoretical surface for indirect prompt injection if the data contains malicious instructions, the skill does not include any triggers or logic that would act on such instructions, and the processing is limited to numeric profiling and summary generation.
Audit Metadata