How we test

Each harness has its own status. A badge on a card follows the harness selected in the header.

Not verified

The install line is documented. No passing run log is stored for that harness. Cursor, Codex, and OpenClaw are not verified. Hermes and Goose are not verified except where a log below says otherwise.

Install tested

The install command ran clean and the page links that log. No bundle is in this state today. The files in runs/2026-11-08 are task runs, not install-only checks.

Run tested

On 2026-11-08 the skill finished its job on the OpenRouter free router. Eight logs ran in Hermes: site health check, meeting actions, first touch draft, review replies, invoice chasers, inbox triage, weekly expenses, and the meeting agenda. Weekly meal plan ran in Goose. Each proof link opens that log file.

Failing

Weekly Topic Brief ran in Hermes and did not keep a reliable source link on each claim, including the reruns. It stays failing and out of the search index.

Needs recheck

A tested or failing status older than 45 days is shown as needs recheck. The stored status is left as it was.