DEDUPEDAuthorization notice. harness2 is dual-use. Only use it on software, sites and systems you own or have explicit permission to test.
Who it is forLegal-tech engineer whose agent drafts antitrust or competition analysis and who wants evidence that prompt and tool edits help
The problemYou cannot tell if rewriting your agent's prompts and guardrails improves legal output without a repeatable before/after run.
The outcomeA judged and reported result for one LAB task, with the adopted harness and reflection notes saved under runs/.
What you need first
- Linux, Python 3.12 and uv for installation
- No credentials needed for the local demo (harness2-demo)
- Model-backed runs need Node.js 20+, the Google Cloud CLI, a Vertex AI project and model IDs in .env, and gcloud application-default login
- LAB needs Pandoc and Poppler; WorkBuddy-Bench needs Docker
- Model-backed benchmark runs incur API costs
Steps
- Install the benchmark: harness2-setup lab
- Check prerequisites: harness2-check --bench lab
- Refine one task: harness2 --bench lab --k 1 --mode parallel --only-domains antitrust-competition --max-tasks 1
- Judge the result: harness2-judge --bench lab --k 1 --mode parallel
- Read the summary: harness2-report --bench lab --k 1 --mode parallel
Prompt for your agent
In google-research/harness2, run the LAB benchmark for one antitrust-competition task with --k 1 --mode parallel, then run harness2-judge and harness2-report with the same flags. Confirm judge and report used identical benchmark, k, mode and seed as the refinement run.
Caveat. Judge and report must use the same benchmark, executor, subset, --k, --mode and seed as the refinement run, and model-backed runs cost real API money, so keep --max-tasks at 1.
About harness2
A framework that lets an AI agent keep improving its own setup (prompts, tools, memory, skills) while the model itself stays unchanged. Part of this week's top 10 repos: ranked #2 on Repo Radar. View google-research/harness2 on GitHub.
More with harness2
- Audit which harness edits were adopted and why Read the adopted harnesses and reflection notes from a run to learn which prompt or tool edits drove a gain.
- Watch an agent improve its own setup in 5 minutes, no API key Run the built-in demo to see both recursion modes edit a harness with real Python executions and zero credentials.
- Boost an office, web or coding agent on WorkBuddy tasks Refine an agent's harness on one WorkBuddy-Bench subset to see if office, web or code results improve.
- Run a clean benchmark with network limits against rubric leaks Lock down network access before evaluating so an agent cannot look up public benchmark answers.
- Compare parallel and sequential refinement on the same task Run the same benchmark task in both recursion modes to pick the cheaper or stronger refinement strategy.
- Run harness refinement with the Codex executor Switch from the default OpenCode executor to Codex to see how the executor changes refined-harness results.
- Resume a long refinement run after it stops midway Re-issue the same command to continue a multi-task run without redoing finished tasks.
Disclaimer
Deduped lists open-source repositories that belong to other people. We do not own, maintain, audit, or vouch for any of them, and a repo appearing here is not an endorsement or a safety check. The use cases, steps, and prompts are written by AI and may be incomplete, outdated, or just wrong. Where a use case is marked "Seen in the wild", we link to public posts where someone described something similar, but we have not verified those posts and the use case itself is still our AI's own write-up. Use all of this at your own risk. Only run code, commands, or agents on systems and accounts you are authorized to use, and read the code and every command before you run it, especially anything that touches credentials, files, or the network. Nothing here is professional, security, legal, or financial advice. All trademarks and project names belong to their respective owners, and their use here does not imply affiliation. Content is provided as is, without warranties of any kind. Found a mistake, a safety problem, or something you want removed? Email [email protected] and we will look at it promptly. Full disclaimer page