Harness² vs Mirror CLI

Harness²

Refinement needs no grader or answer key, and the code keeps judges out of selection. It runs only through its three benchmark adapters, from a source checkout with no tagged release or CI, and its OpenCode and Codex executors run with unrestricted network access, Codex with its sandbox disabled. The benchmark gains are the authors' claims.

Mirror CLI

Mirror CLI is Apache-2.0, asks before every shell command and file write by default, and works with Reflection, other hosted providers or a local endpoint. It is a one-day-old 0.1.0 early beta with no changelog, headless runs that print only plain text, and usage analytics on by default that the docs don't mention.

Harness²

1. Don't plan to call Harness² as a service. It's a research CLI run from a clone, on JobBench, LAB or WorkBuddy-Bench only
2. Ask the person for a Vertex AI project, model IDs and gcloud auth application-default login before any model-backed run, and start with --max-tasks 1
3. Run executors behind an egress firewall and block huggingface.co and github.com after setup, as the docs advise, so rubrics can't be fetched
4. Write .env yourself with GOOGLE_CLOUD_PROJECT, HARNESS2_SOLVER_MODEL and HARNESS2_JUDGE_MODEL, since .env.example isn't shipped, and never commit it
5. Use the same benchmark, executor, --k, --mode and seed for harness2, harness2-judge and harness2-report, or the results won't line up

Mirror CLI

1. Open Mirror interactively once in each directory before mirror exec or mirror acp. Both refuse to run until the usage and folder notices are accepted
2. Pass --approval-mode auto to mirror exec only inside a container or VM. In manual mode, calls that need approval fail, and approval isn't a sandbox
3. Set --max-iterations on unattended runs. Turns are unlimited by default
4. Check the result with your own tests. exec returns 0 whenever it prints an answer
5. Never share logs recorded with --log-level debug. The docs say they contain request headers with the API key