[dmccoystephenson]

Summary

Two stale claims in RESEARCH.md were corrected, both found by a Stage A documentation sweep.

• PR Generalize Phase 4 external-signal step to support no-CI repos #28 entry under §1: CI was described as running only scripts/check_docs.py. The unit tests of that script are also run by .github/workflows/ci.yml, so that scope was added. The conclusion drawn in the entry is unaffected, since CI still does not exercise the template's behavior.
• PR Add template-version header to generated skills for drift detection #33 entry under §8: the version comments were said to go at the top of every generated skill. That was true when the PR shipped, but they are now placed directly below the H1 heading, so a note was added and the historical description was kept.

No tracking issue exists for these corrections; they were found during triage.

Sweep coverage

Every Phase 7 source was checked against create-dev-loop.md, .github/workflows/ci.yml, and scripts/check_docs.py, with CLAUDE.md read for restated conventions. All ship dates cited in RESEARCH.md were also compared with GitHub's mergedAt, and no mismatch was found. The README.md, CONTRIBUTING.md, SECURITY.md, .github/ templates, and scripts/README.md needed no change.

Not covered: RESEARCH.md citation URLs were not re-resolved, and the existing confidence levels were not re-assessed. "Last reviewed" was left unchanged for that reason.

Drift that could not be corrected in this PR was filed:

• Overwrite mode is documented as idempotent, but Steps 6 and 7 are not #172: overwrite mode is documented as idempotent, but the commands in Steps 6 and 7 are not. The Steps are what is wrong, so the fix was left for an implementation cycle.
• Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173: "at the top" is still used in the Step 4 TEMPLATE_VERSION row and at CLAUDE.md:53. Neither file can be changed by an autonomous documentation PR.

Research grounding

No finding applies. Only existing entries were corrected, and the template was not touched. Finding §1's accuracy is preserved, and no finding was shipped by this PR.

Test plan

• python3 scripts/check_docs.py: "Doc consistency check passed."
• python3 -m unittest discover -s tests -v: 42 tests ran, all passing.
• CI scope was confirmed by reading .github/workflows/ci.yml, and the comment placement by reading create-dev-loop.md:124-127.
• Behavioral validation is not applicable: neither create-dev-loop.md nor the template was changed, so no /create-dev-loop run was needed.

Deferred backlog

• CLAUDE.md still links gardener under the old dmccoystephenson owner #106 and half of Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173 require a CLAUDE.md edit, for which separate authorization is required.
• [self-audit] The repo's own validation anchor is structurally unrunnable in a headless dispatch, so template changes can never merge autonomously #109 and Unverified: whether an in-checkout .worktrees/ survives gardener's refresh (git clean -fdx) #114 need a maintainer decision.
• Overwrite mode is documented as idempotent, but Steps 6 and 7 are not #172, the other half of Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173, and every other open template-rule issue change create-dev-loop.md. Each would be handed off UNVERIFIED and would block later cycles while held open, so a documentation sweep that can be merged was preferred.

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, scored against head 6c16067 with doc-consistency green. This review was performed inline in the same session that wrote the change, so it is not independent.

Universal items:

• Scope: PASS. RESEARCH.md is the only file in the diff, with 2 lines changed, and both are the stale claims named in the PR body.
• Tests-new: PASS (not applicable). No function was added.
• Tests-fix: PASS (not applicable). No code bug was fixed, and no check covers RESEARCH.md prose.
• Tests-mutation: PASS (not applicable). This is a Stage A cycle, not Stage B.
• Sibling structure: PASS (not applicable). No file was added.
• Sibling renames: PASS (not applicable). Nothing was renamed.
• Docs: PASS. The Phase 7 table was walked during the sweep. Remaining drift was filed as Overwrite mode is documented as idempotent, but Steps 6 and 7 are not #172 and Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173, and "at the top" was searched for across all *.md files, so the class was covered and not only the instance.
• Issue resolution: PASS (not applicable). No issue is claimed as resolved.
• Closing keywords: PASS. The published body was read for keyword-plus-reference pairs, and none were found. The references to Overwrite mode is documented as idempotent, but Steps 6 and 7 are not #172, Step 4 and CLAUDE.md still place the version comments at the top of the generated skill #173, CLAUDE.md still links gardener under the old dmccoystephenson owner #106, [self-audit] The repo's own validation anchor is structurally unrunnable in a headless dispatch, so template changes can never merge autonomously #109, and Unverified: whether an in-checkout .worktrees/ survives gardener's refresh (git clean -fdx) #114 are all plain mentions.
• CI: PASS. doc-consistency passed on the PR head. Its scope covers the links and parity checks only, not the factual claims edited here.

Repo-specific items:

• Placeholder parity: PASS. check_docs.py passed, and the Step 4 rows were also compared with the template by hand in both directions, since the script checks only one. The two placeholder mentions in the diff are unchanged context.
• README Step parity: PASS (not applicable). No Step was changed.
• Fence escaping: PASS (not applicable). create-dev-loop.md is untouched.
• Phase numbering: PASS (not applicable). create-dev-loop.md is untouched.
• Version comments: PASS. The new note was checked against create-dev-loop.md:124-127.
• Research grounding: PASS. The PR body states that no finding applies. No finding was shipped, so no Implementations entry was added.
• No back-ported specifics: PASS (not applicable). create-dev-loop.md is untouched.
• gardener interface: PASS (not applicable). Steps 3, 5, and 6 are untouched.
• Retrofit note: PASS (not applicable). No rule was promoted.
• Behavioral validation: PASS (not applicable). Neither the Steps nor the template were changed.

Weakest evidence: the PR #33 note attributes the move to #36 on the authority of CLAUDE.md, not from the commit history. The PR #54 title lists #36 in its range, which is consistent with that attribution. No do-not-auto-merge path is matched.

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

---

drafted by Claude on behalf of Daniel Stephenson

[Copilot]

🟢 Approval recommended

There are no active review issues for the documentation-only updates to RESEARCH.md.

No open issues 🎉

What changed in this PR

This documentation-only PR updates two implementation notes in RESEARCH.md to reflect current CI coverage and generated-skill formatting, while preserving historical context.

Changes:

• Clarifies that CI runs both scripts/check_docs.py and its unit tests, but does not validate template behavior.
• Notes that generated skills’ version comments now appear directly below the H1 heading.

File	Description
RESEARCH.md	Corrects CI scope and clarifies current version-comment placement in historical implementation entries.

_{Review effort: Balanced}