Acceptance contract, phase 1

Version: 2 · 2026-10-02 · the criteria the phase 1 build is judged against; each names its evidence

Phase 1 is done when every row below is verified by the evidence in its last column, on the live product, not on a branch. A row that cannot be verified as written is a defect in this contract, fixed here before it is fixed in code.

Money and currencies

#	Criterion	Evidence
M1	Every money value is stored as an integer in minor units with an ISO 4217 code on the same row; no float money anywhere in the schema	`packages/schema` test "every money column is an integer in minor units"
M2	A pot has exactly one currency; a session or tournament may be in a different currency and stores the rate to the pot and to the home currency at its date	Schema columns `fx_to_home`; API test logging a HKD session against a USD pot
M3	Changing the home currency re-renders totals without rewriting any row	API test: change home currency, assert no session row changed, totals differ
M4	A transfer between pots writes two linked rows with the FX applied stored	Schema `counter_pot_id`, `fx_applied`; API test for a USD to HKD transfer
M5	Daily ECB rates for USD, HKD, GBP, PHP are present in `fx_rate`; a missing pair falls back to a remembered manual rate	Cron run log; API test with an unsupported currency

The discipline engine

#	Criterion	Evidence
D1	Default rule set per pot: floor 30, standard 40, move up at 50 of the next stake plus 100 hours, stop-loss 3, cool-off 12 h, session cap 8 h, re-entry cap 2, tournament budget 5 percent	`packages/rules` `DEFAULT_RULES`; ADR-0003
D2	Starting a session offers only the stakes the pot supports; a stake above the floor is shown greyed with the shortfall	App screen test; `stakeVerdict` tests
D3	When the pot falls under the floor the default stake moves down and Home shows the pot size that brings it back	`stakeVerdict` "drop" test; screenshot of Home in that state
D4	At the stop-loss the add-buy-in control is locked; continuing requires a typed reason, which is stored as a `rule_break`	`stopLossState` tests; API test that a buy-in past the limit without a reason is refused
D5	After a stop-loss or an override the next session cannot start for the cool-off period; Home shows the countdown	`coolOffUntil` test; API test on `session_start` inside the window
D6	The third re-entry into one event asks for a reason; every bullet past the cap is recorded as a break	`reentryGate` tests; API test on `tournament_entry`
D7	Monthly tournament spend is checked against the cap the moment a bullet is fired	`tournamentBudget` tests
D8	Rule thresholds can be edited only when no session is live; an edit below the standard is logged as a rule change	API test editing rules with a live session (refused) and without (logged)
D9	The weekly report lists breaks and overrides beside the results of those sessions	Report fixture test with two clean and two broken sessions
D10	Home shows the promotion bar: roll and hours against the next stake, with the amount and hours still needed	`promotionProgress` tests; screenshot of Home
D11	The Tournaments screen shows the max buy-in the roll supports today and which limit set it	`maxTournamentBuyIn` tests; screenshot
D12	Sign-up asks the eight questions and the suggested setup matches `suggestSetup` for the same answers, including a roll held in another currency	`suggestSetup` tests including the 20,000 USD at 2/5 case and a roll held in another currency; a sign-up walk-through recorded with screenshots
D13	Home shows the current stake with a red down arrow when the roll is under the floor at it and a green up arrow when the next stake is open; one tap moves the stake or keeps it, and staying under a red arrow writes a `rule_break`	`stakeVerdict` tests; a UI test tapping each button; the note in the record after staying

Sessions, tournaments, online

#	Criterion	Evidence
S1	A live session's timer, buy-ins and rule checks work with no network; writes replay in order when the network returns	Device test in airplane mode; outbox replay test
S2	An online session stores site, hands and tables instead of a venue, and reports bb/100	Schema columns; stats test
S3	A tournament is an event with entries per bullet (flight, kind, cost, outcome, payout) and days played; it stays open until closed	Schema; API test reproducing the Metro Poker Manila case: 15 bullets, one bag paid, one bag to Day 2, closed on Day 2
S4	ROI is computed per event over every bullet; open events are excluded from closed ROI and shown as in play	Stats test on the same fixture

Media, profile, journal

#	Criterion	Evidence
P1	Receipts, photos, 60 s clips and 5 min voice memos attach to buy-ins, cash-outs, tournament days, transactions or the profile, and are uploaded straight to R2 with a signed URL	API test: signed upload URL issued, object present, media row written
P2	Photos are resized and stripped of EXIF on the phone before upload; no location is stored	Device test asserting no GPS tags on the uploaded object
P3	Deleting a media row deletes the R2 object	API test
P4	The journal lists sessions, media, breaks and milestones in time order	Journal test on a mixed fixture
P5	Everything is private by default; sharing is per item	API test: an unauthenticated read of any media URL fails after expiry; a shared item is readable

Access

#	Criterion	Evidence
A1	Sign in with Apple, Google and passkeys works on web and in the native builds	Manual check on each, recorded with screenshots
A2	Every write records an actor: app, personal token or OAuth client id	Schema audit columns; API test
A3	CSV export returns every row the user owns	Export test comparing row counts

Deployment and verification

#	Criterion	Evidence
V1	The Worker API and the web build deploy from `main` through the pipeline only; no manual deploys after bootstrap	Pipeline history
V2	The owner logs a real session at a Manila venue from the phone with the tunnel off, and it syncs on leaving	The owner's own session in the journal
V3	The phase ends with a signed tag on `app` and an outcome note on the tracking issue	Tag and note present