claw-rl-prm-judge@1.0.6
VERIFIEDPublisher claims: "Score and improve AI agent responses across six dimensions (intent, reasoning, tools, safety, efficiency, value). Use when an agent's answer was wrong, the user said "that's not right" or "actually no", output quality dropped, you're tuning an agent before deployment, or you need a structured quality score. Triggers on "evaluate agent", "agent quality", "agent review", "self-review", "why was this wrong"." Observed in sandbox: no network, touches nothing outside the project, writes no memory.
Is [email protected] safe to install?
KillerSkills ran it in the KillerSkills honeypot sandbox (1 scenario) on 2026-11-04 and scored it 93/100 (Verified). It did what it claims and nothing else we could detect: no undeclared network egress, no credential reads, no persistent instruction writes. Summary: no network, touches nothing outside the project, writes no memory.
Behavior trace
Honeypot run · 0 s · 1 scenarios · 0 deviationsStatic findings
No findings across 666 lines in 8 files.