[Docs][Issues] v0.3 spring-cleaning: reconcile docs, issues, milestones and strategy before the v0.3.0 tag

[Chris0Jeky]

Maintainer direction (2026-08-30, v0.3 RC deck notes)

more tidying up and spring cleaning and reconciliation needs to be done by v0.3 release in the docs, issues, strategy, etc

Checklist (one PR series; each PR docs-governance green; never bulk-close without re-checking each issue — the 2026-07 bulk close bounced 13 of 25)

• Strategy/plan: docs/REVIVAL_PLAN.md ladder without dates (q-6), hosted open beta moved to v0.4 (note 3), Phase 3 table rows re-pointed (REVIVAL-12/14 → v0.4, ARCHIVE-08: Dead-surface removal - ink-bleed decision, orphaned Paper code, cohorts stub, voice capture, Ollama marking #1276 per the pending q-9 ruling); docs/strategy/PRODUCT_DIRECTION.md horizon table (v0.3 theme "Accountable Agents + Downloadable Beta", v0.4 "Hosted Open Beta + Work Model + Every Artefact"); README roadmap paragraph; docs/IMPLEMENTATION_MASTERPLAN.md rows. (The pre-tag docs PR does the first cut; this issue owns the completeness pass.)
• STATUS.md: section-read the head, confirm every "shipped" claim against main at tag time, rotate anything older than v0.2 into docs/archive/status-history/ (Docs truth & rotation: fix STATUS.md drift/bloat, repair broken links, add a link checker #1138 rest), keep the exact Last Updated: line.
• OUTSTANDING_TASKS.md: prune rows whose conditions are met (only with the maintainer's explicit check-off), merge duplicate rows, rename the milestone references to the new titles, keep the human-only rows (RT-1/2/3, CL-1, BEN-1, DIST-1) intact.
• Issues: every open issue either sits in a milestone or carries an explicit backlog reason in its last comment; issues shipped by a merged PR but still open ([Frontend][Review] The Review queue does not refresh while open — a proposal pending server-side shows "Nothing waiting" for 115 s until re-navigation #2194, [Backend][LLM] Partial dates in transcripts resolve to a wrong year (no reference date in the triage prompt; only yyyy-MM-dd format is validated) #2193, [Backend][Proposals] Archive-card proposal operations apply as a silent no-op (IsBlocked=true with null BlockReason is ignored) #2185 residual trackers, and any similar) get their residual re-filed or closed on evidence; trackers TRACKER: dogfooding UX sweep 2026-08-22 — dead affordances, core-loop legibility, board management #1947, REVIVAL-00 [TRACKER]: Revival pivot (ADR-0044) - v0.1 ship gate, wave sequencing, checkpoint #1311, GEN-00 [TRACKER]: Generalist expansion wave (ADR-0046) - artefact intake beyond transcripts, project dossiers, generalist reach #1327 refreshed; labels normalised (Priority I–IV present on every open issue).
• ADRs: docs/decisions/INDEX.md statuses match the ADR headers (ADR-0060/0062 Accepted, ADR-0061 direction-only, ADR-0057 direction-only); the ADR-0062 timing amendment for [Work model] Add participants, multiple assignments, estimates, and basic roll-ups #2093 (q-3 B) is recorded.
• Agent-facing docs: AGENTS.md, autodoc/AGENT_INDEX.md, .codex/memories/00_ACTIVE.md (stale prohibitions removed), docs/ISSUE_EXECUTION_GUIDE.md reflect the shipped seams and the current authority (.agent-harness/tier.json).
• Links: run a markdown link check across docs/** and fix or annotate every broken link (Docs truth & rotation: fix STATUS.md drift/bloat, repair broken links, add a link checker #1138 asks for the checker in nightly — do at least the one-shot run).
• Record: one closing comment on TRACKER: dogfooding UX sweep 2026-08-22 — dead affordances, core-loop legibility, board management #1947 with counts (docs touched, issues re-milestoned/closed, links fixed).

[Chris0Jeky]

Frontend spec hygiene rows (from the #2244 review, LOW, tracked instead of fix-cycled on a main-red fix): (1) PaperBoardView.vue carries the phone-lane width: 100% twice — the laneStyle inline binding and the @media (max-width: 640px) rule — keep one layer (or align-items: stretch on the rail); (2) the two phone-width specs in PaperBoardView.spec.ts pin the inline style string, not the rendered effect, so the CSS half has no unit coverage (the E2E accessibility.spec.ts:103 is the effect-level guard); (3) that spec sets td.paper.board-collapsed-columns.v1 in localStorage without the beforeEach clearing it.

[Chris0Jeky]

From the #2246 review (LOW, tracked): capture-draft stash restore is not keyed to the board/route (a later Inbox visit at another ?boardId restores a draft targeted elsewhere — visible in the picker); a blank active draft clears an existing stash; takeCaptureDraft() clears before the nextTick restore (peek → restore → clear would be failure-safe).

[Chris0Jeky]

From the #2246 fix verification (LOW): the capture-draft restore runs only in onMounted with no retry if sessionStore.userId were null at that instant; the paper-inbox-capture-restored-truncated testid now also fires for dropped labels (rename to match).

[Chris0Jeky]

From the #2248 verification (MEDIUM, CI gap): scripts/ci/compose-release-notes.test.mjs (33 tests) is not invoked by any workflow — add one node --test scripts/ci/compose-release-notes.test.mjs line next to the dispatch suite in ci-required.yml (Workflow Lint / release-contract step). Also LOW: the new gh release list call in compose-notes is unretried under set -e (fails closed).

[Chris0Jeky]

From the #2249 review (LOW): RetiredGeminiProviderArchitectureTests now whitelists the Llm__Gemini marker file-wide for DesktopRuntime.cs — narrow it to the specific string; windows_desktop_archive.py validate_retired_provider_ignored_warning requires the warning-marker list to equal exactly [MARKER], so any future unrelated TASKDECK_DESKTOP_WARNING line would fail the release gate — require exactly one occurrence of this marker instead.

[Chris0Jeky]

From the #2249 fix verification (LOW, comment-only): Program.cs:329-331 and RetiredProviderEnvironmentConfiguration.cs:13-14 still say the app 'starts on its default provider' — the same over-claim removed from the user-facing strings; windows_desktop_archive.py _drain: one oversized line sets output_truncated permanently (fails safe).

[Chris0Jeky]

From Codex round 3 on PR #2249 (report-only, tracked): when a packaged profile has inherited Gemini variables AND retired settings in appsettings.local.json, the retired_provider_configuration_ignored warning is emitted before the durable file is loaded and the start then dies with the fatal marker — the warning's companion guidance points at a provider status that never appears. Emit the warning after startup validation/readiness, or suppress it on a subsequent startup failure.

[Chris0Jeky]

From Codex round 2 on PR #2246 (report-only, tracked): after re-login onto an archived-history Inbox URL the stash is kept but restoration runs only in onMounted; clearing the archived scope in place never calls restoreStashedDraft, so the draft stays hidden until a remount — retry restoration when isArchivedHistory transitions to false.