[shiny-code-app]

Prepared website settings could leave a qualified release waiting for Chris to click Ask the Client now, because only preview-review evidence opened automatic requests. Changed prepared settings now use the existing invitation worker and receipts, including settings-only releases on the same build. The Client reviews the prepared choices on the release page and still decides whether to accept; this creates no acceptance or promotion authority.

The settings digest participates in the candidate/cache identity, with one compatibility reader for historical four-field markers. Completion cannot settle a new settings request against an earlier settings snapshot. Adoption records the same settings receipt, so lost responses, restarts, adoption and re-preparing identical settings on another artifact avoid duplicate notifications. Stale, withdrawn, unchanged, unverified and blocked intent supplies no automatic opening. Pure engineering candidates retain their existing behavior.

Closes #4702. Parent: direction#3 CM website live and used, under Q759 and brief LPASK-1024.

The Q759 drill prerequisite is deployed: running API and Odoo worker were independently read on43e5901a44659007baf76f6aa6b11f21d6fb6777, whose history contains launchplane#4689's landing55b0981f. The hold is satisfied; full current-head CI still gates Ready. CMSET-1024 owns the prepared CM settings; this PR changes no product build or settings and does not edit either neighbor's files.

Validation: three failing-first publisher cases; reproduced adoption extra ping and persistent-cache omission before their fixes. All 195 tests across full affected invitation, release-review, HTTP, website-settings and Client-release modules pass. Whole-repository Ruff/format, mypy (1760 files), role words and diff checks pass. Release notification/acceptance safety classification; no authorization, secret, credential or recovery implementation changes. JetBrains inspection NOT RUN because the brief prohibits launching or touching any JetBrains process. Full current-head CI is required before Ready.

Review: own Anthropic claude-opus-5-5[1m] review; disposition and final review will be posted as a head-bound comment. Adoption, same-build cache and withdrawn-settings false completion defects reproduced and fixed. Fresh design checks and the final outside review found no blocking defect. Settings-only action wording corrected and literal wording assertions removed. House style forbids a replacement docstring. Prepared-settings behavior is documented in its owning page; CMLIVE's release-review document is left untouched under the brief's ownership boundary.

Removed: the Director's manual send step for qualified settings releases; no new route, operation or grant. Net lines: +306 (+339/-33), of which +246 are tests, +14 documentation and +46 runtime.

Historical limitation: four-field legacy receipts cannot establish which settings were announced before this deploy. The existing reviewed admin adoption remains an attestation about the current checklist. There is no existing request for the CM candidate, per CMSET-1024; no guessed migration or manual Client message is added. A settings-only withdrawal preserves the historical request while the review page offers no pending release.

[shiny-code-app]

Review result: weighed
Head: 178a80b

Review of record weighed for head178a80b3bc0e0a11dba4bd39b6c909be5345389b.

Anthropic claude-opus-5-5[1m], CLI-reported, through my own review_with_model runs. After two rounds I stopped patching for fresh read-only design checks; the final complete candidate/legacy compatibility design also received Anthropic review. Final review reports no blocking defect and agrees with both DIRECTION files.

Acted on, with failing-first proof: missing settings automatic openings; adoption's duplicate settings ping; same-build worker cache suppressing changed intent; withdrawn settings being falsely settled by an earlier release. Legacy adoption/completion and existing renewed-request/reminder behavior are preserved. Mixed releases retain both actions and silently remove withdrawn settings. Literal wording assertions were removed; tests use source notes or fixture inputs. All195 affected whole-module tests and broad static gates pass.

Weighed legacy-only observations: older four-field receipts have no settings identity, so they cannot prove which settings were announced. No guessed migration is added. Existing reviewed admin adoption deliberately attests the selected manual request against the current checklist; that authority is unchanged and grants no acceptance. A hypothetical previously opened legacy settings request could retain its old ambiguity or receive its first structured settings announcement. CMSET-1024 confirms no request exists for the current CM candidate, so no such duplicate is being sent here. These are historical-data limits, not defects in newly emitted exact-settings receipts; legacy automatic settlement retains its established compatibility behavior. Legacy-only time filtering would not establish settings identity and would conflict with intended same-version renewal semantics.

Docstring replacement is declined under the no-new-docstring house rule; the obsolete preview-only statement was removed. Prepared-settings behavior is documented on its owning page; CMLIVE's release-review.md is reserved by the brief. Inspection NOT RUN under the explicit IDE ban. No unresolved blocking finding remains against this head.

[shiny-code-app]

Current-head CI attempt1: the changed behavior's local qualification remains green. Shard6 failed in the unchanged test_github_request_timing worker-command test when its mocked run-once subprocess exceeded30seconds. That focused test passes locally in4.025seconds. The test has no diff from main and does not exercise the invitation publisher. Two PostgreSQL checks and browser-evidence verification also reported cancellation; available PostgreSQL logs end in ordinary cleanup, and the cancelled browser job's log blob is unavailable. No product assertion failure was established for those cancellations.

The first CI round is waiting only on ci-gate to become terminal. Then LPASK-1024 will use the maintained failed-job rerun on this same reviewed head, under the three-retry budget. No timeout, test, source or runner mutation is being used to bypass the gate. Own runner logs confirmed actual shard execution despite an empty GitHub step projection. Q759's drill deployment prerequisite is satisfied by the running API/worker revision43e5901a containing55b0981f.

[shiny-code-app]

Ready for the merge train

Head: 178a80b

[shiny-code-app]

Launchplane completed the merge-train step for this pull request.

• Repository: cbusillo/launchplane
• Base branch: main
• Pull request: Automatically request Client review for prepared website settings #4705
• Controller action: land_batch
• Controller record: merge-train-batch-landing-plan-20261024T181954Z-9d71364b07d6d476

Launchplane landed this pull request through protected batch PR #4731.

Launchplane manages this comment and will update it as the train moves.