evals-live-run

Orchestrate repeatable live-device skill evaluations and controlled debugging sessions.

2|Updated Feb 17, 2026
One-click install
npx skills add https://github.com/clawperator/clawperator --skill evals-live-run
Or copy as Structured Prompt for Agent
Please help me install this Agent Skill.
Skill: evals-live-run
Source: https://github.com/clawperator/clawperator/tree/main/.agents/skills/evals-live-run
Command: npx skills add https://github.com/clawperator/clawperator --skill evals-live-run

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill includes scripts (resource) and references (resource) components.

What problem does it solve?

The evals-live-run skill provides a convenient entrypoint for repeatable live-device skill proving runs, consolidating harnessed evaluation flows and retained-log workflows.

Core Features & Use Cases

  • Orchestrates Solax orchestrated-cold-start eval runs against real devices for repeatable results.
  • Provides a wrapper to run Pack A Samsung android-version benchmarks and other eval scenarios with explicit device targeting.
  • Enables debugging of retained logs and selective replay of newest eval batches.

Quick Start

Run the Solax cold-start eval on a real device using the included helper scripts to kick off an end-to-end evaluation.

Frequently Asked Questions about evals-live-run

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I run repeatable live-device skill evaluations on real hardware?

Run repeatable live-device skill evaluations by using an eval harness that accepts device serials and operator packages to orchestrate controlled debugging sessions. This process targets real devices to produce consistent artifacts and logs for analysis.

What is retained-log debugging and how does it work during live-device evaluations?

Retained-log debugging is a workflow that enables selective replay of the newest evaluation batches. It works by retaining logs produced during live-device runs, allowing operators to debug issues using the preserved artifacts after the evaluation completes.

Can I target specific devices when running Solax cold-start proving evaluations?

Yes, you can target specific devices when running Solax cold-start proving evaluations. The evaluation harness accepts explicit device serials, ensuring the orchestrated cold-start runs execute against your intended real hardware.

Does the live-device evaluation harness support Android version benchmarks for Samsung devices?

Yes, the live-device evaluation harness supports Android version benchmarks for Samsung devices. It provides a wrapper to run Pack A Samsung benchmarks and other evaluation scenarios with explicit device targeting for accurate results.

Do I need specific operator packages to start a live-device debugging session?

Yes, you need specific operator packages to start a live-device debugging session. The evaluation harness requires these packages along with device serials to properly orchestrate repeatable skill proving runs across real devices.

What limitations exist when replaying retained logs from previous evaluation batches?

A limitation when replaying retained logs is that selective replay is restricted to the newest evaluation batches. You cannot replay arbitrary historical logs; the system focuses on recent live-device debugging sessions for targeted analysis.