Readiness

Readiness answers three questions with evidence instead of opinion: is the platform safe to run on paper with people watching, is it production quality, and is it good to trade real money. Behind each answer is one list of checks. Every check is proven by something that runs: a test suite, an eval score against its bar, a screen check, an audit finding being closed, a live reading from the running system, or a dated sign-off that expires. A check nobody has proven yet counts against the answer, on purpose. The page never rounds up.

The house owner opens it from Admin › Readiness. The operator of a book opens Readiness of your book and sees only the checks that are measured per book, for that book; the platform-wide list stays with the owner.

The three answers

The top of the page carries three cards: Safe on paper, Production quality, Real money. Each says Yes or Not yet, how many of the checks it needs are proven, and the first few things to clear, worst first. The answers are cumulative: production quality needs everything safe-on-paper needs, and real money needs everything. Click a blocker to jump to its row; click "Show its checks" to narrow the list to what that answer needs.

Real money is the answer the go-live procedure reads. It does not replace the paper track record scorecard, which is one of its checks.

What a row tells you

Each row is one check: a plain title, one line saying what stands between it and proven, and which answer first needs it. The colours mean:

Open a row for the full wording, the owner, whether it is measured per book or once for the platform, the earlier judgment if there was one, and Details: the exact proofs with their numbers, the command that ran, and where to look. That is the one place on the page where engineering words appear, because there the thing shown is the identifier.

By area

Nine tiles group the checks by the question they answer: data, strategy and risk, orders and stops, agents, security, operations, product finish, edge over doing it yourself, and legal. Each tile shows how many of its checks are proven; pick one to narrow the list, pick it again to clear.

What changed, and what was caught by eye

Two cards sit under the answers for the owner. What changed since the last run lists checks that became proven and checks that slipped back, so a morning look takes seconds. Caught by eye counts the things a person noticed that no check caught, this week and over the last eight weeks, with a target of zero. Every such catch is logged with the check that would have caught it; the card lists the ones whose check does not exist yet, and each new catch adds a placeholder row to the list so the fix cannot forget its check.

Run now

Run now (owner only) takes the live readings again immediately: quote freshness and feed, which model each role actually ran, spend against budget, the universe, the paper track record verdict, and any sign-off that has expired since. It is recorded in the audit log and finishes in a second or two. It does not re-run test suites, evals or screen checks; those come from the nightly full run, and the page says which kind of run the report came from. If a refresh cannot finish, the last-run line says so and the previous report stays on screen.

Where the list comes from

The list itself lives in the repository next to the code, one entry per check with its proofs, and the same command that feeds this page can be run there. Adding a check is a one-minute edit; the nightly run picks it up. Nothing on this page changes how the desk trades: it reads, it never steers.