nemo-gym-reward-profiling
Pass
Audited by Gen Agent Trust Hub on May 20, 2026
Risk Level: SAFE
Full Analysis
- [COMMAND_EXECUTION]: The skill outlines the use of CLI tools such as ng_run, ng_collect_rollouts, and ng_reward_profile for managing model servers and generating profiling metrics.
- [CREDENTIALS_UNSAFE]: Command examples use environment variables (e.g., $POLICY_ENDPOINT_KEY) for API keys, which is a secure practice for handling sensitive credentials in scripts.
- [SAFE]: No evidence of prompt injection, malicious obfuscation, or unauthorized data exfiltration was found. The instructions are consistent with the documented purpose of the reward profiling workflow.
Audit Metadata