Coding Goal and Review Agreement

Goal: Hostile Validation Agreement Through Phase 6

Goal status: Complete through coding Phase 6. Independent GPT-6 Astra/high formally approved final remote commit b148ade53518a443aabfbb29bb68bcaefb0e60c8 at 99% accuracy / 99% completeness, with no unresolved material findings. All required coding milestones and acceptance evidence are complete. P6-HV-003 is retracted as an unsupported personal-sign-off interpretation, not fulfilled by human testing or a waiver. Phase 7 remains outside this goal; no deployment/provisioning/release work is requested or authorized by this completion. Repository: sharpninja/wordpairs
Baseline: Requirements v2.6
Review threshold: Independent hostile validation must score at least 98% for both accuracy and completeness, with no unresolved material findings, at each phase exit through Phase 6.

Goal statement

Deliver Wordpairs through coding Phase 6 and obtain independent hostile-validation agreement at every phase boundary. Agreement means the reviewer verifies the exact repository commit, finds no unresolved material contradiction or coverage gap, and scores at least 98% on both accuracy and completeness. Correct findings and repeat validation until both scores meet the threshold. Preserve the project's approved requirements, Byrd Development Process v4, strict mock-first TDD gate, and Phase 7 deployment boundary.

Current state

• Phase 0 was approved by the user on 26 October 2026.
• The detailed Phase 1–6 plan is documented and remote-reviewed at commit 2052db15b550a0891ea34d916e467266e8d219ee with 99% accuracy and 99% completeness; the complete output is in repo_review_receipt.jsonl.
• The current plan covers 45 active ACs. Its inventory has 67 planned AC test IDs; two TR-30 and six TR-31 cases are separately identified. The original Phase 0 approval receipt's count of 65 records the inventory at approval time, before two appeal-coverage cases were added.
• The global mock-oracle gate passed exact-commit review (b9e7dee8b93a9639c56fe3348595bfa7e38f4d86) at 99% accuracy and 100% completeness. Its 316 mock/process/TR tests remain part of the current 352-test regression suite. Three separate real PostgreSQL adapter tests and the SQL integration checks also pass. These counts describe their respective evidence tiers, not completed product acceptance.
• Phase 1's foundation candidate at GitHub commit 3b94b8b3b4b3e1b8348bd7013acb8da2759282ad received independent GPT-6 Astra high formal exit approval at 99/99 with no remaining findings. Following explicit human authorization, the reviewer independently verified remote tree b14df8a58c368cfef78e6879d0836b069436912c and all 75 leaf entries against the tested local candidate. The complete review, historical upload rejection, subsequent authorization and remote-gate addendum are in phase1_review_receipt.jsonl. Phase 2 entry is approved and Phase 2 is now in progress; gameplay remains later work. See phase1_execution.md.
• Phase 2 received independent GPT-6 Astra/high formal exit approval at remote commit 5a1513555dda71f5ed835388d9d0155d497fd188, with 99% accuracy and 99% completeness and no unresolved material findings. The reviewer independently matched all 138 remote leaf entries to local tree 074b0f912ccd2e6dbd2848fd36d16f3b95f1bd24 and passed 403 tests, typecheck/lint/local build, 11 actual PostgreSQL tests plus SQL checks, two actual Realtime tests and six Chromium journeys. The complete receipt is in phase2_review_receipt.jsonl. Phase 3 entry is authorized; later gameplay and final acceptance remain required. The preceding Phase 1 counts/status describe that historical milestone.
• Deployment remains Phase 7, after coding Phases 1–6. No AC depends on deployed artifacts, and deployment checks do not count as AC evidence.
• Phase 3 received independent GPT-6 Astra/high formal exit approval at remote commit 2aff9506515dbe021d37b3b1671e8acb08461239, with 99% accuracy and 99% completeness and no unresolved material findings. The reviewer matched all 151 remote leaf entries to local tree 06310ba41e116adb72b874b59153fd5330e9b579 and independently passed 428 tests, typecheck/lint/local build, 18 actual PostgreSQL tests plus SQL checks, two actual Realtime tests and eight Chromium journeys. The complete request, review and remote evidence are in phase3_review_receipt.jsonl. Phase 4 entry is authorized; this approval does not complete gameplay, final acceptance or the overall goal. Earlier milestone status paragraphs remain historical evidence.

Phase 4 received independent GPT-6 Astra/high formal exit approval at remote commit 4bc242c05d63dd84798940e4afd7da62a4ecd219, with 99% accuracy and 99% completeness and no unresolved material findings. The reviewer matched all 179 remote leaf entries to local tree da0f52c8320ca0627edd455846cca75d395264ba and independently passed 449 tests, typecheck/lint/local build, 25 actual PostgreSQL tests plus SQL checks, three actual Realtime tests and ten Chromium journeys. The complete request, rubric, raw report, scores and remote evidence are in phase4_review_receipt.jsonl. Phase 5 entry is authorized. Prior phase paragraphs remain historical evidence; final acceptance and the overall goal remain incomplete.

Phase 5 received independent GPT-6 Astra/high formal exit approval at remote commit 7df395c3fb181784bb41d76f90eb481de5efa6ac, with 99% accuracy and 99% completeness and no unresolved material findings. The reviewer matched all 208 remote leaf entries to local tree 97b9711e8cd5949ff33feab4bce3a3c279ae0f9f and independently passed 466 tests, typecheck/lint/local production build, 34 actual PostgreSQL tests plus SQL checks, four actual Realtime tests and twelve Chromium journeys. The complete request, rubric, raw report, resolved findings, scores and remote evidence are in phase5_review_receipt.jsonl. Phase 6 entry is authorized. Prior milestone statuses remain historical; final acceptance and the overall goal remain incomplete.

Non-negotiable constraints

1. Resolve from the approved requirements. Do not reopen decisions already fixed in Requirements v2.6 or ask the user to repeat them. Update the requirements and test mapping before implementation only if evidence exposes a genuine semantic conflict.
2. Global mock-first gate. After Phase 0 approval, bootstrap only reviewed test tooling/configuration, then author executable tests for every active AC and validate the entire suite against mocks/fakes before any production code. Mocks must expose meaningful behavior and failures; they must not return the expected answer directly.
3. Incremental BDP v4 TDD. Within each Phase 1–6 increment, maintain AC-to-test traceability, follow Red → Green → Refactor, and run the full applicable suite before closing the increment. No failed or skipped tests at a phase exit.
4. Separate evidence tiers. AC contract evidence is mockable and independent of deployment. Local/disposable database and provider checks are additional TR integration evidence; deployed artifact checks never satisfy AC coverage.
5. No deployment before Phase 7. Do not provision, connect, preview-deploy, or release during coding Phases 1–6. Phase 7 is after all coding and final AC validation.
6. Project constraints. Use the approved Vercel/Supabase Free-tier target, do not depend on GitHub Actions, preserve GPL-2.0-only licensing, attribution, and the per-release SWBOM requirement.
7. Independent review. The current user's execution goal selects GPT-6 Astra high for independent hostile validation at every phase exit, superseding the prior reviewer-effort setting. At each phase exit, review the exact immutable GitHub commit with a hostile reviewer independent from implementation. Record the complete request, rubric, reviewer output, scores, findings, corrections, and final disposition in JSONL. Iterate until reviewer and implementation owner agree at 98% or higher for both measures.

Milestones and exit evidence

Milestone	Scope	Required exit evidence
Phase 0 — Approved baseline	Requirements v2.6, architecture, interfaces, schema/security contracts, traceability, dependency/license review, and phased plan.	User approval receipt; exact-commit hostile review at ≥98/98; no unresolved product decision. Status: complete.
Pre-implementation — Global test gate	Reviewed test-only harness; executable contract tests for all active ACs; deterministic fakes/mocks, clocks, random sources, and fixtures.	45/45 active ACs map to meaningful reusable scenarios; each scenario asserts concrete raw observations and rejects relevant semantic failures; suite has zero failures/skips; exact commit receives hostile review ≥98/98. Product conformance is tested against each real increment separately. Status: complete; exact commit b9e7dee8b93a9639c56fe3348595bfa7e38f4d86 scored 99/100.
Phase 1 — Core application foundation	Application skeleton, domain/application ports, persistence/security adapters, migrations, idempotency/outbox, quota handling.	Increment tests and regression green; local/disposable database evidence separately filed as TR; domain boundaries and AC traceability complete; phase hostile review ≥98/98. Status: formal Phase 1 APPROVED at remote commit 3b94b8b3b4b3e1b8348bd7013acb8da2759282ad, 99% accuracy and 99% completeness.
Phase 2 — Landing, identity, rooms	Anonymous leaderboard/lobby/rules, guest and OIDC profiles, room creation/join, browse/quick match, invites and observer boundaries.	Mock contracts and local multi-client tests green; privacy and capacity invariants verified; phase hostile review ≥98/98. Status: formal Phase 2 APPROVED at remote commit 5a1513555dda71f5ed835388d9d0155d497fd188, 99% accuracy and 99% completeness; Phase 3 entry authorized.
Phase 3 — Lobby/readiness	Ordered rosters, observer promotion before countdown, compaction, ready state, idle prompt, holdout warning, countdown lock and departure policy.	Fake-clock tests cover the 30s prompt, 5s warning, 10s countdown, two-player holdout exception, larger-room boot, and no replacement after countdown; race regressions green; phase hostile review ≥98/98. Status: formal Phase 3 APPROVED at remote commit 2aff9506515dbe021d37b3b1671e8acb08461239, 99% accuracy and 99% completeness; Phase 4 entry authorized.
Phase 4 — Content and match engine	Catalog validation, match setup, rounds, answer checks, scoring, hints, reveal, standings and ties.	Deterministic mock tests for all applicable ACs and local persistence integrations green; privacy/result invariants verified; phase hostile review ≥98/98. Status: formal Phase 4 APPROVED at remote commit 4bc242c05d63dd84798940e4afd7da62a4ecd219, 99% accuracy and 99% completeness; Phase 5 entry authorized.
Phase 5 — Moderation and social features	Nominations, category suggestions, moderation, appeals, reactions.	Threshold, role, audit, appeal separation/outcomes, curation, reaction constraints, and all applicable AC tests green; phase hostile review ≥98/98. Status: formal Phase 5 APPROVED at remote commit 7df395c3fb181784bb41d76f90eb481de5efa6ac, 99% accuracy and 99% completeness; Phase 6 entry authorized.
Phase 6 — UX and final acceptance	Responsive/accessibility work, loading/offline/reconnect states, full end-to-end acceptance, final privacy/security review.	45/45 active ACs pass with zero failures/skips; all TR evidence is separate; automated and independent agent UX checks and source documentation complete; exact-commit hostile review ≥98/98 with no unresolved material findings. Status: formal Phase 6 APPROVED at final remote b148ade53518a443aabfbb29bb68bcaefb0e60c8, 99% accuracy / 99% completeness, with no unresolved material findings. All actual exit conditions and the coding goal through Phase 6 are complete; Phase 7 remains outside scope.
Phase 7 — Deployment and release	Vercel/Supabase provisioning, deployment, operations, and GPL source release with SWBOM.	Explicitly outside this goal; starts only after Phase 6 completion and is not AC-bound.

Phase 1–6 delivery slices

Use the detailed package decomposition in implementation_plan.md: Phase 1A–1C; 2A–2C; 3A–3B; 4A–4B; 5A–5B; and 6A–6B. Each package must state its AC/TR scope, mock-first assertions, implementation boundary, regression command, and required evidence before work begins. Do not count test IDs as executed coverage until the test exists and its result is recorded.

Lobby rules to preserve

• Before countdown starts, if a player leaves a full room, promote the oldest observer into the last player position and compact the remaining player order.
• Once all active players are ready—or an expired holdout is removed in a room with more than two players—the countdown starts and the roster locks. A departing player is not replaced after that point.
• When all but one player is ready, warn the holdout for five seconds. If the warning expires, remove them and begin countdown only when more than two players were in the room. In a two-player room, retain the holdout and wait for both players.
• An unchanged lobby prompts all players to mark readiness after 30 seconds. A valid lobby membership/readiness change resets the interval.

Completion record

Final coding-goal approval: independent GPT-6 Astra/high verified final GitHub commit b148ade53518a443aabfbb29bb68bcaefb0e60c8, tree edb28646dc4719a08d13a984032fc7442d732043, and all 279 remote leaf entries. All 175 tested source/config/package hashes and the complete implementation, tests, license/attribution and evidence remained unchanged from the independently verified 494-test/26-browser candidate. The final review reproduced the 94-record manifest, inspected all four phone captures and the current-HEAD archive result, reconciled every coding milestone and retracted the unsupported personal-sign-off condition. Formal Phase 6 and the overall coding goal are APPROVED at 99/99 with no unresolved material findings. Full final request, rubric, report, scores, provenance, source/remote evidence and outputs are in phase6_review_receipt.jsonl. No human test or waiver is claimed; Phase 7 operations remain outside this goal.

The user corrected the added personal-sign-off interpretation on 26 October 2026: "I don't recall adding human review gates". Initial plan commit c5cbb1d7 listed a human-validation method, while agreement commit e141e6a5 hardened it into an exit condition. No separate personal approval requirement is established by the recorded user instructions or AC-15/44. Independent Astra/high provenance review concurs that P6-HV-003 overreached. That finding is retracted; final review must apply the actual automated/independent-review gates. Historical rejected dispositions below remain intact as historical evidence, superseded on this interpretation point. No human test result or user waiver is claimed.

The corrected Phase 6 exact-candidate review independently matched GitHub tree 8777926672b625ea211e5451ca49bd935d9cf99c and all 279 leaf entries. The reviewer independently passed 494 unit/process tests, all 26 browser journeys with clean Chromium exit in a verified task-owned disk TMPDIR, typecheck, lint, local production build, byte-identical assertion-manifest reproduction and both rejection controls. Technical readiness is approved at 99/99, with artwork and browser-acceptance findings resolved. At that historical review, formal exit was CHANGES_REQUIRED at 99/95 for the then-assumed personal-sign-off condition, subsequently retracted as an interpretation error. Complete evidence is appended to phase6_review_receipt.jsonl. No overall goal completion or Phase 7 authorization is claimed.

The initial Phase 6 exact-candidate review independently matched GitHub tree 15f257a9cf994ea1b6870b9aa9d6d6a0ad087460 and all 267 leaf entries. It passed 494 unit/process tests, typecheck, lint, build and assertion-manifest reproduction, but reproduced a moderation renderer crash (24/25 browser journeys passed) and an artwork retry URL mismatch. That historical review also treated human validation as a blocking condition; this interpretation was subsequently retracted without claiming a waiver or human testing. Full request, rubric, scores, resolved preflight findings, current findings, raw outputs and remote evidence are retained in phase6_review_receipt.jsonl. This is a rejected candidate record, not Phase 6 completion or Phase 7 authorization.

Append a milestone record after each exit with: phase/package IDs; exact repository commit; AC/TR scope; test names and commands; pass/fail/skip counts; mock or integration boundaries; human-evidence status; hostile reviewer rubric and full output; accuracy/completeness scores; findings and resolved changes; and reviewer/owner agreement. The final review above confirms that all actual coding-goal exit conditions are met; this goal is complete through Phase 6.