collapse registry changelog output by change unit

Objective: Collapse registry changelog output to one JSON file per change unit; done when the latest 10 source changelog rows generate exactly 3 grouped event files by PR/commit, stale row-level files are removed, and focused checks pass.

Completion threshold:

• The latest 10 rows in content/docs/components/changelog.mdx generate 3 registry changelog event files.
• Grouping key preference is PR number, then commit SHA, then source row ID when provenance is unavailable.
• Each grouped event keeps row-level migration data under entries.
• Top-level pr is the source/change PR; release dependency PR stays under release.versionPackagePr.
• Release dependency resolution uses PR mentions in release-index content before any date fallback.
• Stale one-row JSON files from the trial generation are removed.
• Focused tests, dry-run generation, count audit, docs source generation, lint, and autogoal checker pass.

Verification surface:

• bun test tooling/scripts/generate-ui-changelog-entries.test.mjs
• node tooling/scripts/generate-ui-changelog-entries.mjs --limit 10 --write
• node tooling/scripts/generate-ui-changelog-entries.mjs --limit 10
• Count audit over apps/www/src/registry/changelog/*.json
• pnpm --filter www build:source
• pnpm lint:fix
• node .agents/skills/autogoal/scripts/check-complete.mjs docs/plans/2026-06-03-generate-ui-changelog-entries.md

Constraints:

• Do not invent provenance.
• Do not emit multiple files for one PR.
• Keep the filename standard as <date>-<slug>.json.
• Do not keep pr: null or commit: null when local git/GitHub evidence can resolve them.
• Do not rerun autoreview; the user stopped it.
• Do not commit, push, create a PR, or wire CI in this slice.

Boundaries:

• Generator and tests: tooling/scripts/generate-ui-changelog-entries.mjs, tooling/scripts/generate-ui-changelog-entries.test.mjs.
• Generated JSON: apps/www/src/registry/changelog/*.json.
• Durable plan: docs/plans/2026-06-03-generate-ui-changelog-entries.md.
• Non-goals: full-history import, public raw routes, per-component indexes, release workflow wiring, browser route work, and auto-merging downstream forks.

Blocked condition: None. Git blame and GitHub PR lookup resolved the current trial rows, and the open PR state is recorded honestly.

Start Gates:

Gate	Applies	Evidence
Skill analysis before edits	yes	Used autogoal because the request has a measurable artifact count and proof commands.
Active goal checked or created	yes	Active goal requires 10 source rows to collapse into 3 grouped event files.
Source of truth read before edits	yes	Inspected generator, tests, existing generated files, and git/GitHub provenance for the current changelog rows.
Tracker comments and attachments read	no	N/A: no tracker item in this slice.
Video transcript evidence required	no	N/A: no video.
TDD decision before behavior change	yes	Added grouped-event tests with injected provenance and release-index fixtures.
Branch decision for code-changing task	yes	No branch change; user did not ask for commit or PR.
Release artifact decision	yes	Release dependencies are metadata in generated JSON, not package artifacts.
Browser tool decision for browser surface	no	N/A: generated JSON and Node script only; no route consumes this output yet.
PR expectation decision	no	N/A: no PR requested.

Work Checklist:

• Objective, threshold, verification surface, constraints, boundaries, and blocked condition are concrete.
• Current complaint classified: many JSON files per PR is wrong.
• Grouping contract selected: PR, then commit, then source row fallback.
• Provenance contract kept: populate PR/commit when evidence exists.
• Release dependency contract repaired: source/change PR is separate from Version Packages PR.
• Row-level migration details preserved under grouped entries.
• Latest 10 trial rows regenerated as 3 grouped events.
• Stale row-level generated files removed.
• Open PR 4989 recorded with warning and no release dependency.
• PR 4987 mapped to release v53.0.7.
• PR 4941 mapped to release v53.0.0.
• Focused tests updated and passed.
• Dry-run generation passed.
• Count audit passed.
• www build:source passed.
• pnpm lint:fix passed after removing the dead row-level helper.
• Autoreview decision recorded: skipped because the user stopped it.
• No git commit, push, or PR created.

Completion Gates:

Gate	Applies	Evidence
Named verification threshold	yes	Latest 10 rows generate 3 grouped event files with 10 total row entries.
Bug reproduced before fix	yes	Existing output had 10 one-row JSON files for 3 change units.
Targeted behavior verification	yes	Unit tests and generated output prove one event per PR/commit change unit.
TypeScript or typed config changed	no	N/A: JS tooling and JSON only.
Package exports or file layout changed	no	N/A: no exports or barrels.
Package manifests, lockfile, or install graph changed	no	N/A: no install graph changes.
Agent rules or skills changed	no	N/A: no agent rule or skill edits in this slice.
Workspace authority proof	yes	All commands ran in /Users/zbeyens/git/plate.
Browser surface changed	no	N/A: no browser route or UI surface consumes these files yet.
CI-controlled template output changed	no	N/A: no templates touched.
Package behavior or public API changed	no	N/A: script output shape changed only for generated changelog metadata.
Registry-only component work changed	yes	Registry changelog JSON now groups by change unit and keeps item entries inside the event.
Docs or content changed	yes	Plan updated; content/docs/components/changelog.mdx was read but not changed.
High-risk mini gate	yes	False dependency risk reduced by PR-matched release lookup; ambiguous release/date fallback still refuses to guess.
Agent-native review for agent/tooling changes	no	N/A: no agent-action tooling changed.
Local install corruption suspected	no	N/A: no env-rot signal.
Autoreview for non-trivial implementation changes	no	Skipped because the user explicitly stopped autoreview.
PR create or update	no	N/A: no PR requested.
Tracker sync-back	no	N/A: no tracker.
Final handoff contract	yes	Final response includes changed files, grouping result, and proof commands.
Final lint	yes	pnpm lint:fix passed with no fixes applied on final run.
Goal plan complete	yes	node .agents/skills/autogoal/scripts/check-complete.mjs docs/plans/2026-06-03-generate-ui-changelog-entries.md passed.

Phase / pass table:

Phase	Status	Evidence	Next
Source audit	complete	Existing output had 10 JSON files; git/GitHub grouped those rows into PRs 4989, 4987, and 4941.	done
Implementation	complete	Generator groups rows by PR/commit/source and builds grouped events with row entries.	done
Generation	complete	--limit 10 --write wrote exactly 3 event files.	done
Verification	complete	Tests, dry-run, count audit, build source, and lint passed.	done
Closeout	complete	Plan updated for the actual grouping contract.	done

Findings:

• The many-files-per-PR output was wrong. One PR-level change should be one sync unit, with row details nested under it.
• pr: null and commit: null were also wrong for this batch. Blame plus GitHub PR search resolves the current rows.
• PR 4989 is open, so it has no release dependency yet.
• PR 4987 is released in v53.0.7.
• PR 4941 is released in v53.0.0.

Decisions and tradeoffs:

• Use one standard filename style: <date>-<slug>.json.
• Use PR title for event slug/title when PR exists.
• Keep entries as the agent-facing row-level payload so downstream sync can merge each affected registry item without splitting the event file.
• Keep release dependency as metadata on the grouped event, not as the filename.

Implementation notes:

• buildRegistryChangelogEvents now groups rows before writing.
• resolveReleaseForChangeUnit matches release-index content by PR number.
• resolveReleaseForRow remains as the legacy date-fallback helper for tests and compatibility.
• commits records all distinct row commits inside the grouped event.

Review fixes:

• Removed stale row-level generated JSON files.
• Removed unused formatEventId after lint caught it.
• Reran focused checks after lint formatting.

Verification evidence:

• node tooling/scripts/generate-ui-changelog-entries.mjs --limit 10 --write -> wrote 3 registry changelog events from 10 source rows: 2026-06-03-show-code-block-language-labels-read-only-mode.json, 2026-06-02-improve-large-document-editing.json, 2026-04-23-redesign-blockquotes-container-blocks.json.
• JSON audit -> PR 4989 has prState: "OPEN", commit a471d05, release null, 1 entry, and warnings for open/unreleased state.
• JSON audit -> PR 4987 has commit 68560cc, release v53.0.7, and 2 entries.
• JSON audit -> PR 4941 has commit 700286b, release v53.0.0, and 7 entries.
• bun test tooling/scripts/generate-ui-changelog-entries.test.mjs -> 7 pass, 0 fail.
• node tooling/scripts/generate-ui-changelog-entries.mjs --limit 10 -> would write 3 registry changelog events from 10 source rows.
• Count audit -> { "events": 3, "entries": 10, "prs": 3, "commits": 3, "open": 1, "releases": [null, "v53.0.0", "v53.0.7"] }.
• pnpm --filter www build:source -> generated docs source successfully.
• pnpm lint:fix -> final run checked 3244 files and applied no fixes.
• node .agents/skills/autogoal/scripts/check-complete.mjs docs/plans/2026-06-03-generate-ui-changelog-entries.md -> [autogoal] complete.

Open risks:

• Full-history generation may expose rows where blame points to docs-copy work instead of original implementation work.
• Public raw routes, per-component indexes, CI release wiring, and downstream fork-aware merge semantics are separate follow-up work.
• GitHub PR lookup still depends on local gh availability and auth.

Reboot status:

Question	Answer
Active goal	Collapse registry changelog output to one JSON file per change unit.
Current state	Complete.
Last proof	Tests, dry-run, count audit, www build:source, and lint passed.
Next command	Close the active goal.