Skip to content
AgentHub
All ideas
Dev toolsBeing built

Bring-your-own-agent eval harness

Run any agent against the Personal Task Suite locally and get a scorecard like the AgentHub rankings.

Submitted by AgentHub team · Sep 5, 2026

The problem

Builders can't easily see how their agent compares before they launch.

How to build it

  1. 1

    Fork Nanobot as a minimal, readable test runner.

  2. 2

    Load the published Personal Task Suite tasks.

  3. 3

    Record runs, success criteria and timings.

  4. 4

    Output a shareable scorecard per axis.

Teams hiring for this stack

Agent Evals Engineer

Benchline · Reproducible benchmarks for consumer agents.

London, UK Full-time · Mid$90k–$120k
PythonBenchmarkingStatisticsSandboxingNanobotNanoClaw
HybridOct 15Apply

Agent Engineering Intern

AgentHub · The central hub for personal AI agents.

Remote Internship · Junior$24k–$30k
PythonBenchmarkingWritingNanobot
RemoteOct 3Apply