docs: record §10 onboarding bullet in ADR 0023; trim changelog

[vpciii]

Closes the governance gap the PR #25 review correctly flagged: the v0.12.0 §10 onboarding bullet is a normative practice extension that shipped without an ADR — against the methodology's own "record the change in an ADR" / "arguing for a practice → ADR" rules, and unlike comparable additions (ADR 0014, ADR 0016).

• New ADR 0023 records the §10 one-command-onboarding bullet as an accepted extension of ADR 0003's Dependencies practice (ADR 0003 is append-only, so it's recorded anew, not edited). §10's decision trail is now ADR 0003 + 0023. Marked Proposed — accepted on merge.
• CHANGELOG trim — drops the 0.12.0 entry's #22/#24 process backstory the review flagged as too journal-y; keeps the substance + adoption line and cites ADR 0023.

The honest note: this gap was caught by the same-model automated reviewer that the cross-model adversarial process and I both missed — which is the answer to "should we turn that reviewer off": no.

🤖 Generated with Claude Code

[claude]

Claude finished @vpciii's task in 2m 7s —— View job

---

Review — ADR 0023 + changelog trim

• Read PR diff
• Read repo standards (methodology.md §10 / "Changing this document", ADR 0003, ADR format, all 22 prior ADR statuses)
• Check ADR numbering, format, cross-references
• Verify changelog ↔ ADR consistency
• Post review

This is a clean, well-reasoned governance PR. It correctly closes the gap (a normative §10 bullet that shipped without an ADR), follows the house ADR format exactly (Context → Decision → Alternatives → Consequences → Adoption impact → References → Nygard footer), and every cross-reference checks out: §10 does have exactly the three ADR-0003 bullets plus the new fourth one (methodology.md:419-433); the two quoted "Changing this document" rules are verbatim-accurate (methodology.md:531-532, 541-542); ADR 0003 is correctly treated as append-only (recorded anew, not edited); and the CHANGELOG now cites ADR 0023 with bullet wording that matches the doc.

One substantive issue.

1. Status: Proposed is inconsistent with every prior ADR — and won't get flipped (adr/0023-...md:3)

ADR 0023 is the only ADR of the 23 marked Proposed; all 22 of 0001–0022 are Accepted. The PR body says "Marked Proposed — accepted on merge," but there's no convention or process step in this repo that performs that flip — every prior ADR simply landed as Accepted in its introducing PR. As written, once this merges the file will sit permanently as Proposed (a follow-up to change it would be the only way), which is exactly the doc/decision drift this ADR exists to prevent.

It's also arguably already inaccurate: the bullet shipped in 0.12.0 and is live in both methodology.md and the CHANGELOG, so the decision is de facto accepted, not proposed.

Recommend setting it to Accepted to match the established convention (a status change is metadata, not an edit to the decision, so it doesn't conflict with append-only).

Fix this →

Nit (non-blocking)

• Decision blockquote is a paraphrase, not the verbatim §10 bullet (adr/0023-...md:31-36). The recorded quote reframes the text ("no tooling mandate, consistent with ADR 0003's 'weigh and record'") rather than reproducing methodology.md:426-431 ("Reproducible builds aren't reproducible if the setup isn't…"). The substance is faithful, so this is fine — just noting that a reader treating the ADR as the source of the exact wording would find it differs from the doc.

Nothing else flagged — numbering is correct and unique, no spec/test obligations apply to a docs-only change, and the commit follows Conventional Commits.