Is Agent Reward Bench Safe?

Agent Reward Bench — Nerq Trust Score 51.2/100 (C- grade). Score based on 2 independent trust signals. Last analyzed: 2026-03-20

Agent Reward Bench is a Python package with a Nerq Trust Score of 51.2/100 (C-), based on 3 independent data dimensions. Last analyzed: 2026-03-20 Security: 90/100. Popularity: 15/100. Data sourced from PyPI registry, GitHub repository, NVD, OSV.dev, and OpenSSF Scorecard. Last updated: 2026-03-20. Machine-readable data (JSON).

Safety Guide: Agent Reward Bench

What is Agent Reward Bench?

Agent Reward Bench is a Python package — Official library for AgentRewardBench: Evaluating Automatic Evaluations of Web Agent Trajectories.

How to Verify Safety

Run pip audit or safety check. Review on PyPI for download stats.

You can also check the trust score via API: GET /v1/preflight?target=agent-reward-bench

Key Safety Concerns for Python package

When evaluating any Python package, watch for: dependency vulnerabilities, malicious uploads, maintenance status.

Measured Signals

Agent Reward Bench has a Nerq Trust Score of 51/100 (C-). This score is a composite of automated measurements of security, maintenance, community, and quality signals.

Key Takeaways

• Agent Reward Bench has a Trust Score of 51/100 (C-).
• The score is a measured composite — it is not a suitability judgment. Evaluate the individual signals against your own requirements.
• Query the current measured values via the Nerq API.

Disclaimer: Nerq trust scores are automated measurements based on publicly available signals. They are not endorsements, verdicts, or guarantees of suitability. Always evaluate the signals against your own requirements.