Cline vs Harness²

Cline

Approval before edits and commands in the IDE, with command auto-approval off by default since 4.0.0. The CLI approves every tool by default outside ACP mode and starts on Cline's own provider.

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.

Cline

1. Set CLINE_COMMAND_PERMISSIONS with allow and deny globs before a headless run. The CLI approves every tool by default
2. Pick a provider with -P and a key, or run cline auth. The default provider needs a Cline sign-in
3. Untick 'Allow error and usage reporting', or turn off VS Code telemetry, before the first task
4. Pin the CLI version. 2.3.0 was a malicious publish
5. Keep the hub on 127.0.0.1 or set ROOM_SECRET, and run 3.0.30 or later

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