Prepare v0.4 feature contracts and QA outcome matrices

[Chris0Jeky]

Taskdeck needs a concrete v0.4 qualification backlog and an honest guide to current versus planned behavior. Add a source-linked feature catalogue, eleven session matrices with 44 initial cases, a reusable session template and a CSV outcome ledger. All future outcomes start as NOT RUN; experimental, unconfigured, prototype/stub and deferred states are distinguished.

Refs #2898. Refs #2899.

The tracker and eleven child issues are assigned to v0.4 and synchronized to Project Pending / Priority III. Existing performance, visual baseline, hosted implementation, dogfooding and release owners remain linked; their milestones and release authority are unchanged. The capability guide records the main baseline through #2895 and the separate #2897 delivery status.

Validation: documentation links pass across 689 Markdown files; GitHub operations governance and diff checks pass; all 44 CSV case IDs are unique and initialized NOT RUN. The QA change is documentation-only relative to the already-reviewed #2897 reminder fix included for integration. Merge #2897 first; this PR then contains seven QA documentation files only. No new QA session or release acceptance is claimed. Bounded independent documentation review and hosted checks are recorded on this PR.

Integration head 6b78ccd includes #2897 head b57f7fa. Backend, frontend, workflow and tooling content match that reviewed delivery exactly. Both competing top-of-file documentation additions were retained; QA content is unchanged. This allows combined qualification to run while #2897 completes.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-10T12:01:34.432409Z	4bd2794	New commits

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 589578300a

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Chris0Jeky]

Independent fresh-context Luna documentation review: CLEAN, no causal HIGH/CRITICAL findings. Reviewed the seven-file, 500-line QA diff at 5895783 against b344a5a. Cited numeric contracts match source; all 44 case IDs are unique, four per issue2899-2909, allv0.4/NOT RUN with blank candidate/actual fields. Existing release gates, review/apply authority and human acceptance are preserved.

Nonblocking coverage note: explicitly include #2089 capture links/disposition lifecycle during session expansion. Tracked with concrete expected behavior on #2900 and #2907; no extra review/fix loop or claim of completed testing.

Current integration head6b78ccda9 retains exactly that QA diff relative to b57f7fa. The two STATUS/MASTERPLAN additions were retained side by side. Backend/frontend/workflow/tooling content matches independently reviewed #2897; the documentation review remains valid. Local689-file links, operations governance, CSV schema/count/status inspection and diff checks pass. Hosted qualification remains pending.

[Chris0Jeky]

One factual correction pass is committed at 4bd2794. The guide and CORE cases now distinguish UI/API proposal apply, CLI direct card mutations (existing #1131 hardening), and MCP proposal get/list/dismiss with no approve/apply operation. Audio limits distinguish the browser's 60-second recording timer from size/MIME-bounded uploads, which have no server duration check. The reachable cohort dashboard is explicitly classified Stubbed with its endpoint and #1142/#1276 owners.

The incorrect cross-interface authority statement was a material false safety claim; the audio and cohort comments were factual product/QA contract corrections within the requested catalogue work. All were reconciled in one small pass, with no runtime changes. Linked issues #2899/#2900/#2905, plan and CSV are synchronized; 44 unique NOT RUN cases remain. Documentation links689/governance/diff and CSV inspection pass. A single scoped correction verification is pending; no broader review cycle is requested.

[Chris0Jeky]

Scoped correction verification is CLEAN at4bd2794d1202d8a034736a6a0e82f120ebc406a9 against6b78ccda9. The independent reviewer checked CardsCommandHandler, ProposalTools, AudioAnswerRecorder/ThinkingAudioService, feature flags/router/AutomationMetricsController and the CSV. All three corrections match source; 44 unique cases remain NOT RUN with blank candidate/actual fields. No further findings or runtime changes. This completes the bounded review and one correction round; hosted run34473893755 remains the exact-head gate.

[chatgpt-codex-connector]

💡 Codex Review

Here are some automated review suggestions for this pull request.

Reviewed commit: 4bd2794d12

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

Codex can also answer questions or update the PR. Try commenting "@codex address that feedback".

[Chris0Jeky]

Parent #2897 merged to main as beace64 after required run34472353566 passed. Main's tree exactly equals parentb57f7fa5b, which this QA branch already contains.

Revalidated #2910 against actual main: merge base remainsb57f7fa5b; the diff is seven QA documentation files (514insertions). Backend/frontend/workflow/tooling diffs are empty. The reviewed documentation plus scoped factual corrections remain unchanged, so the existing review covers the resulting base and combined input tree. Current head4bd2794d1 still awaits required run34473893755. No receipt-only commit, retarget or new test claim is introduced. Future QA issues remain open; #2808 implementation is independently complete.

[Chris0Jeky]

Review triage complete at head 884871a.

• The three P2 threads from the 5895783 pass (interface authority, 60-second audio scope, cohort stub) were already answered by the 4bd2794 correction commit and independently re-verified CLEAN; they are now resolved as outdated.
• The remaining live P2 ("Count confirmation within the 50-version cap") was confirmed against source and is a real documentation defect: ThinkingAudioService.WriteAsync refuses a Save once ListByCaptureAsync(...).Count >= MaximumWrittenVersions - 1 (49) and that list counts every representation for the capture, so provisional transcriptions share the pool and ConfirmAsync occupies the reserved 50th slot. The boundary row would have had QA report a conforming refusal as a failure. Fixed in 884871a (one line, documentation only).

node scripts/check-docs-governance.mjs and node scripts/check-doc-links.mjs (689 Markdown files, 0 broken relative links) pass at the new head. The fix stays inside the reviewed finding and changes no runtime behaviour, so no further review round is owed; the hosted required run at 884871a is the merge gate. Codex review credits are exhausted, so no new connector pass is expected on this head.