[dmccoystephenson]

Summary

Every RESEARCH.md Implementations entry was checked against GitHub's recorded merge date (gh pr list --state merged --json number,mergedAt). Seventeen of the twenty cited PRs matched. Three did not:

Entry	Recorded	mergedAt (UTC)
PR #67 (Tests-fix CI-revert fallback ladder)	2026-07-26	2026-07-27T01:50:04Z
PR #136 (checkout-form Tests-fix revert), cited under two findings	2026-08-25	2026-08-27T02:19:17Z
PR #162 (Stage B mutation requirement)	2026-10-19	2026-10-20T15:06:58Z

All four lines are corrected. The #136 and #162 dates were wrong in any timezone. The #67 date was right in US time but not in UTC, and the rest of the file uses UTC (for example, "PR #65 (2026-07-27)" for a merge at 2026-07-27T00:49Z). So the "How to use this document" section now says that "the date shipped" means the PR's merge date in UTC, as GitHub records it in mergedAt. That makes the convention checkable.

These are observed instances of #128: the ship date in each entry was written before the PR merged and was never updated. This PR does not resolve #128. The question there is a format choice for the maintainer (record the authoring date, or defer the ship date as pending), and CLAUDE.md restates the format, so it cannot be changed autonomously. #128 stays open.

No tracking issue: the drift was found during triage. Related: #128.

Deferred issues

The other open issues were not taken this cycle. Nearly all of them change the generated template's behavior, which needs a manual /create-dev-loop validation run that a headless dispatch cannot do (#109). Others need a CLAUDE.md edit, which requires separate authorization (#106, and half of #128). A fully verifiable docs-only fix was picked instead.

Research grounding

No finding's substance changes. Only the recorded dates in existing Implementations entries are corrected.

Doc sync check

No Step, placeholder, or template text is touched. CLAUDE.md's "the date shipped" wording is unchanged and still consistent with the clarified definition.

Test plan

• python3 scripts/check_docs.py — "Doc consistency check passed."
• python3 -m unittest discover -s tests — Ran 41 tests, OK
• Each corrected date checked against mergedAt from gh pr list --state merged
• No /create-dev-loop run is needed: this change is docs-only and does not touch the template (UNVERIFIED-not-applicable).

This PR description was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

🤖 Generated with Claude Code

---

drafted by Claude on behalf of Daniel Stephenson

[dmccoystephenson]

Self-review rubric (posted as a plain comment; the review was performed inline during a headless dispatch):

• Scope: PASS. Only RESEARCH.md is modified: four date corrections plus one clause defining "the date shipped". The clause is needed so the Add Tests-fix CI-revert fallback ladder and scratch-file guidance #67 correction can be checked; without it, that date is ambiguous between US time and UTC.
• Tests-new: N/A. No code changed.
• Tests-fix: N/A. This is a docs correction, not a code bug fix. The regression evidence is the mergedAt comparison quoted in the PR body.
• Sibling structure: PASS. No new files.
• Sibling renames: PASS. Nothing renamed. The Make the checkout form Phase 4's default Tests-fix revert and gate ladder rung 1 #136 entry appears under two findings, and both copies were corrected.
• Docs: PASS. CLAUDE.md asks for "the date shipped", which the clarified definition is consistent with. CONTRIBUTING.md and the PR template don't restate the date format. No other source of truth is affected.
• Issue resolution: PASS. There is no Closes line, and RESEARCH.md Implementations entries record a ship date that is unknowable when the entry is written #128 is referenced as related only. Its format question is left to the maintainer.
• CI: PASS. doc-consistency passed on the PR head.
• Placeholder parity / README Step parity / Fence escaping / Phase numbering / Version comments: N/A. create-dev-loop.md is untouched.
• Research grounding: PASS. The body states that no finding's substance changes.
• No back-ported specifics / gardener interface / Retrofit note: N/A. No template or Step change.
• Behavioral validation: UNVERIFIED-not-applicable. The change is docs-only.

Judgment call for a human: the UTC definition is a small convention addition in a file that CLAUDE.md points to. It does not pick any of #128's options, but a maintainer who prefers local-time dates may want to revert that clause and the #67 change. The #136 and #162 corrections are wrong in any timezone, so they hold either way.

Of the 20 cited PRs, 17 already matched mergedAt. All were checked in one gh pr list --state merged --json number,mergedAt call.

This review comment was drafted during a Gardener session (https://github.com/Stephenson-Software/gardener).

---

drafted by Claude on behalf of Daniel Stephenson