operator-eval

What problem does it solve?

The hidden operator-only evaluation loop over local runtime/eval artifacts can be difficult to run safely and reproducibly within standard user workflows.

Core Features & Use Cases

• Automates local evaluation tasks by reading, validating, and executing a single operator request from the local surface.
• Supports regression review, replayable evaluation runs, candidate promotion, and baseline freezing while preserving provenance.
• Ensures isolation by keeping evaluation artifacts under runtime/eval and separating runtime activity logs from normal user data.

Quick Start

Ask the agent to run the operator-eval workflow for the current evaluation request and return results to the main agent.