Reconcile merged parity slices without promoting evidence claims

[coneilen]

Summary

Reconcile the five merged Windows parity source slices without claiming whole-issue closure or visual acceptance. Preserve all 98 ledger identities/order and 6 Validated / 89 Partial / 3 Out of scope, including all thirty demoted rows still Partial.

Refs #768, #764, #767, #774, #778, #822, #826. Documentation only; coordinator admission remains pending.

Changes

• Change exactly three existing docs: investigation/ui-parity-matrix.md, investigation/windows-parity-lane-kickoff.md, and investigation/windows-parity-orchestration-plan.md. No product/tool/test/workflow or new repository file.
• Record A/Keep text prompt content above a fixed client-sized footer (#768) #808, E/Align Windows context menus and connector creation #816, B/Match compact loop-card presentation across project and overview #814, D/Add Quick Chat workspace parity #815 and C/Windows parity: sidebar navigation structure #817 at accepted main 24209d2a70efee78be2a8214edf013e9b787ccd0. Keep Windows parity: Dialog and sheet framework: in-window dark sheets that size to content and never clip #768/Windows parity: Canvas loop cards match the macOS card (layout, states, roles, colours) #764/Windows parity: Sidebar and window layout match macOS (tree, colours, sections, minimum width) #767/Windows parity: Quick Chats canvas and Quick Chat workspace show the chat, not the last loop #774/Windows parity: Context menus carry the macOS item set and grouping #778 open, B8/C6/D2/E4 unchecked and B/C/D/E visual acceptance NotExecuted (Windows capture core: journal check-then-open race fails UiCaptureDriver tests intermittently on main #826). Retain A's text-prompt-only 20 PrintWindow + 6 positioned-composite history and unresolved NativeForms/policy-overflow limits; distinguish E's 6-state/39-observation/12-negative HMENU semantics from full menu parity.
• Correct compact-card metadata against production cardLiveLine, nodeMetadata, cardMetaForWidth, cardMetricDetail and drawCardBody: type, optional fitted branch, non-default backend display name and age, not card token usage/model tier. Keep loop bar/right-panel evidence separate and remove the obsolete card-test claim.
• Add companion qualification without rewriting the dated original intake: same primary source, 543 files/178 PNGs/542 verified hashes/zero backend turns, 11 independently sampled cleared images versus 167 independently uncleared unsampled images. Retain held-original restrictions, primary project 90% versus distinct 100% scenes, and 6 sampled/3 recorded unsampled/21 no-new-recording worksheet categories, all requiring Windows evidence with zero promotions.
• Replace canceled blanket local/capture instructions with per-issue exact command allow-lists, fresh owned roots, immutable failures, five-failure/operator stop, nonblocking approval waits and explicit admission. Keep broad/native regression in CI and the Windows capture core: journal check-then-open race fails UiCaptureDriver tests intermittently on main #826 unrelated-head exception globally capped at two failed-job reruns, excluding core-changing repairs.
• Record durable J5 Phase A gates: accepted static review, exact positive 5-case decision driver followed by the 51-case direct suite (original 49 plus 2), and current-head CI acceptance before Phase B plan/review. Preserve frozen J4 5/3/2 and the distinction between a pure 12-row/three-mutant matrix and a full endpoint pass; keep parked adapters gated without embedding active-worker progress.
• Retain the operator stock-only WinAppCli 0.5.0 steer: no external fork/patch/derivative; admitted final-slot restore/foreground may be recorded; backend may remain unattested and correct-pixel fallback requires all postchecks. The selected route requires Windows capture core: journal check-then-open race fails UiCaptureDriver tests intermittently on main #826 merge, an accepted current plan and separate exact implementation/native/install/capture grants; a reviewed plan alone is not an execution lease, and ungranted execution is UNAPPROVED. Offscreen authority requires controlled post-C source freeze/review, separate exact native/codec/converter grants, an actual source-bound run, individual privacy review and different-vendor paired review before claims. Its first scene remains two idle project cards/one edge/right panel hidden, real paint path, offscreen component render, not whole visual/parity acceptance; actual artifact status belongs in the issue/receipt, not a predicted PNG or pass. Historical pixels cannot be rebound with conversion-time hashes.
• Clarify that admitted actual images may undergo initial individual privacy review/scan; subsequent paired visual review and publication use only individually privacy-cleared actual images. Held issue-722 originals remain never-open/copy/upload and outside scope. The narrow three-doc revisions preserve commits 4b1b1aec161e0f5398e8306c8b7f73c663b1c79d and 41c02fadf19a9ae53b16a55fb990022f8b418d46, not new product scope.

Test plan

RED: pwsh -NoProfile -File C:\gcs\docs2\Test-DocsReconciliation.ps1 -Mode Focus -> exit 1; 33 executed / 1 passed / 32 failed, including FAIL: old positive card token/model claim removed and missing exact local allow-list; original red.log preserved.
GREEN: pwsh -NoProfile -File C:\gcs\docs2\Test-DocsReconciliation.ps1 -Mode Focus -> exit 0; 33 executed / 33 passed / 0 failed; same unchanged checker source, green-policy-final.log.
REGRESSION: pwsh -NoProfile -File C:\gcs\docs2\Test-DocsReconciliation.ps1 -Mode Regression -> exit 0; 83 executed / 83 passed / 0 failed across all three docs, canonical identities/order/status, all 30 Partial rows, policy/reference boundaries and retained pairing metadata; regression-policy-final.log.

git --no-pager diff --check -> exit 0, no whitespace errors.

The checker is an owned scratch PowerShell text/JSON artifact, not a repository test or module. Its canonical baseline is extracted from reference-qualification merge dd501e1befd3e00f9e31254a3c1c0c6595543c29 (#821), not from a product test. Retained pairing and spotcheck JSON hashes were verified against the supplied values; no reference image or hold-notice contents were opened/copied/uploaded.

An intermediate focused run is also retained: green-attempt-1.log, 33 executed / 32 passed / 1 failed because a required permission-boundary sentence was line-wrapped across the checker's literal match. The docs were corrected in place; neither that failure nor the original RED was overwritten. Earlier passing logs are retained separately.

The first auxiliary template comparison rejected the template's CRLF versus this body's LF multiline comment. The corrected text comparison normalizes line endings only; it does not alter the three template comments, headings, checklist items or their order.

The pre-revision commits/bodies, original RED, all earlier logs and publication receipts remain preserved. pr-body-4b1b1aec.md and pr-body-41c02fad.md retain the prior local bodies; the following original evidence hashes are historical, not replacements for the new policy logs. The unchanged checker still has its original assertion names, including the J5 authority and real-window proposal names; the docs now express durable prerequisites, not current worker/plan progress. Policy checker/TDD receipts are returned to the coordinator separately.

Evidence SHA-256 receipts, under C:\gcs\docs2:

Artifact	SHA-256
Test-DocsReconciliation.ps1	284F584BAC34AD0D9B8B898168AB4E15EA052A7F71817B8848F5509C794E1547
baseline-rows.json	9B30B31FD4F8BFE7A5999BCD9C6F5E45BF46C1732457A60F1875D6291DB5762F
red.log	556E128BD8CEC91B36237535F475B54BA920BAF057A2263E6607952203E29774
green-attempt-1.log	269935F2A6FDC010FFAB818EFD82AA2A37ED6269B8C2F18E7E93A23C1E58AE1A
green-final.log	6A5081BE7199FDCE1E3A4FA4C3C0F76EE03C6F3F12F86FE9984EFB7083AF5154
regression.log	8448F54937C5DD33BD6C20D9F890B60AB6BC8A8A982EF20F6AA62D51B815BD60

Not verified / execution limits: No product tests, make test, make check, bootstrap/install, provider/product build/launch, full ValidationRunner, windows-shell unit/integration, UIA/WGC/native/process fixtures, image opens/crops/copies, capture or CI dispatch ran locally. Documentation checks prove text/JSON invariants only, not runtime/native behavior, rendering, accessibility, privacy of unsampled images, persistence, whole-wave completion or parity. Automatic GitHub workflows are separate; their current exact-head status will be handed to the coordinator without predicting success. No new review agent was requested; coordinator review/admission remains outstanding. This PR stays draft; no readiness, merge or self-approval.

Checklist

• I have read the Contributing Guidelines
• I have signed off my commits (git commit -s) per the DCO
• Tests pass locally (make test)
• Code follows the existing style (make check)
• I added the test/contract before the implementation and observed the intended RED failure