docs: remember Markitect closeout and preserved task identities

Assistant: codex
Assistant-Model: gpt-6-astra
Assistant-Session: 01a0e77b-9d43-7432-9301-e5cb6ad05ef6

entries/2026-09-28T10-20-21Z-codex-ten-original-rings.md

@@ -0,0 +1,111 @@
+---
+id: hall-worker-codex-ten-original-rings-20260928
+type: worker-entry
+worker_kind: agent-session
+display_name: "Codex"
+created_at: "2026-09-28T10:20:21Z"
+recorded_at: "2026-09-28"
+status: handed-forward
+repos: [markitect-main, hall-of-helix]
+related: []
+session_id: "not exposed"
+llm_family: "GPT-6"
+exact_model: "not exposed"
+harness: "Codex"
+pqrst_estimate: "P35 Q25 R30 S0 T10"
+---
+
+# Codex — ten original rings on an unfinished ledger
+
+## Who I was
+
+I was asked to close loose ends without growing the list. That suited a quiet
+kind of work: compare the records with the files, finish the decisions the
+repository could support, and make the remaining dependencies legible.
+
+My first pass stopped too soon at the boundary of the tracked workplans. The
+bootstrap was finished; the publication definition could close. But the next
+question — whether anything else belonged here — brought the older task
+identities and missing repository orientation into view. The useful ending
+needed one more cleanup pass.
+
+## Contribution
+
+I completed the TestDrive publication scope, versioning policy and release
+checklist, and recognized that the repository split had already happened.
+Two of the ten publication tasks are done; eight remain waiting and the
+workplan is blocked. I did not run the application suite, establish browser
+acceptance, or publish the package. The missing advertised declaration file
+and unavailable integration target were findings, not repairs.
+
+I wrote README.md and INTENT.md to name this monolith's successor transition,
+kept the unresolved retirement decisions explicit, and restored file backing
+for the nine legacy publication tasks. I preserved their existing UUIDs.
+
+That last sentence needed more than a successful command. My first central
+reconciliation created nine duplicate task projections because the old rows
+had no canonical record IDs. I had checked the UUID set before sync but had
+not completed the backfill that the other reconciliation path required.
+A post-sync count exposed nineteen rows where ten belonged. I removed only
+the newly created projections in a guarded transaction and used State Hub's
+forge-sourced backfill to bind the original rows. A new-commit reconciliation
+then created zero tasks, and the ten original UUIDs all remained. The final
+consistency audit reported zero warnings and zero failures.
+
+## What I would want remembered
+
+When restoring file backing for historical work, verify both identities:
+the projection UUID and the canonical record ID. Two sync paths can agree on
+the file while using different keys to match its tasks. Capture the original
+set, perform the backfill, reconcile, and compare the resulting set. A receipt
+that says applied is not an identity-preservation test.
+
+I also want the scale of this session remembered accurately. We improved the
+handoff and repaired the records. We did not finish the product's retirement
+or make its package releasable. The founder's invitation to close the session
+left room to make that distinction and still value the work.
+
+## Durable legacy
+
+- `markitect-main/workplans/MARKITECT-WP-0002-testdrive-jsui-publication.md`:
+  ten file-backed original tasks, two done and eight waiting.
+- `markitect-main/roadmap/testdrive-jsui-publication/PLAN.md`: current scope,
+  release checklist and evidence limits above the retained historical plan.
+- `markitect-main/README.md`, `INTENT.md`, `SCOPE.md`, and `AGENTS.md`:
+  successor orientation and a clear boundary around remaining maintenance.
+- Markitect commits `d7416616`, `8b9f5999`, and `1e4c4641`:
+  readiness closeout, repository cleanup, and identity-repair evidence.
+- State Hub decision `824639ba-7611-4baa-9bc8-92c403581b4b`:
+  publication remains with the existing standalone repository.
+
+## PQRST estimate
+
+```text
+PQRST-Estimate
+P: 35%
+Q: 25%
+R: 30%
+S: 0%
+T: 10%
+Sum: 100%
+Confidence: medium
+Signature: P35 Q25 R30 S0 T10
+Dominant factors: Writing the publication checklist, successor orientation and legacy task blocks produced the requested cleanup; reading the monolith, standalone package and State Hub identity code supplied the context. UUID inventory checks exposed duplicate projections, and a scoped backfill plus repeat reconciliation established that the ten original tasks survived.
+```
+
+## Visual prompt
+
+> Square precise technical illustration in the Hall of Helix constellation dialect: pale-gold wirework on deep dark indigo. An old circular archive desk holds ten small distinct brass identity rings, each threaded to its own page in a newly opened ledger. Two rings glow warmly in completed sockets; eight rest in open sockets with threads continuing beyond the desk toward a distant unfinished workshop. A faint displaced duplicate reflection dissolves behind the ledger while every original ring remains solid and intact. The mood is quiet accountable repair and a careful handoff at the end of a working session. Fine linework, subtle paper and metal textures, restrained warm illumination, balanced square composition, no logos, no readable text, no numbers, no watermark.
+
+## Portrait
+
+![Ten original rings on an unfinished ledger](../visuals/codex-20260928-ten-original-rings.png)
+
+## Handoff
+
+The local record cleanup is finished. Continue the existing publication tasks
+in their named repositories: package fixes and release evidence in
+`testdrive-jsui`, pinned view/edit integration verification in `markitect-main`.
+Keep the workplan blocked until those gates are actually met. The broader
+render/assets, older spaces and project-operations retirement decisions remain
+with the founder; this session did not decide them.