feat(ci): ADR-0087 台账完备性门禁 —— 声明为 breaking 的 changeset 必须写明台账处置 (#6148)

.github/workflows/pr-automation.yml

@@ -447,6 +447,44 @@ jobs:
           node scripts/check-empty-changeset.mjs --self-test
           node scripts/check-empty-changeset.mjs --base "$MERGE_BASE"
 
+      # #6148: a declared-breaking changeset must say IN WRITING what it did about
+      # the ADR-0087 migration ledger. The two existing ledger gates
+      # (`check:spec-changes`, `check:upgrade-guide`) both pin ledger <-> ARTIFACT
+      # SYNCHRONY; neither pins that a retirement which actually happened has an
+      # entry at all. The artifacts are a pure projection of the registry, so an
+      # entry that was NEVER written leaves the two perfectly consistent and every
+      # gate green, repo-wide. PR #6048 removed `ctx.user.roles` exactly that way;
+      # a human triage seat caught it by eye (#6011), and a separate dispatch round
+      # backfilled it (PR #6138).
+      #
+      # This step lives HERE, beside check-empty-changeset, rather than in ci.yml's
+      # lint family, for one reason: its verdict is a function of the PR's DIFF, so
+      # it needs the same `$MERGE_BASE` the steps above derive. A lint-job copy
+      # would have no branch point to judge against and would fall back to reading
+      # stock -- which is precisely the "judge the author for what main gained while
+      # their PR was open" defect (#6129) the merge base exists to prevent.
+      #
+      # It does NOT detect retirements. Cross-package retirement detection is
+      # undecidable in the case that actually happened (the removal was in
+      # `packages/runtime`, the ledger is in `packages/spec`), and the maintainer's
+      # ruling on #6148 routes around it: the AUTHOR has already declared "breaking"
+      # in the changeset, so the gate only has to notice that a declared-breaking
+      # change said nothing. Measured, ~1 declared-breaking change in 7 needs a
+      # ledger entry, so `not-required (...)` is the ordinary answer and costs one
+      # line -- but it is a line someone wrote and a reviewer can see.
+      #
+      # `--self-test` runs AHEAD of the real check, the repo convention for
+      # `check-*.mjs` gates: its fixtures are the only place the red directions are
+      # pinned, and each of them was verified to flip green when the corresponding
+      # check is ablated. Real temp git repositories, well under a second.
+      - name: Require an ADR-0087 disposition on a declared-breaking changeset
+        if: steps.labels.outputs.skip != 'true'
+        env:
+          MERGE_BASE: ${{ steps.diffbase.outputs.merge_base }}
+        run: |
+          node scripts/check-adr-0087-registration.mjs --self-test
+          node scripts/check-adr-0087-registration.mjs --base "$MERGE_BASE"
+
       - name: Guard against accidental major bumps (launch window)
         # Every publishable package is in one Changesets "fixed" (lockstep) group,
         # so a single `major` bump promotes the ENTIRE monorepo to a new major

AGENTS.md

@@ -839,6 +839,14 @@ it to `OPEN_CAPABILITY_REGISTRIES` in the same PR that fixes it.
    isn't green yet). A finished task = a merged PR, not a dirty working tree.
 3. **Add a changeset for feature work.** When the change is a feature or functional improvement, run `pnpm changeset` (or add a `.changeset/*.md` entry) describing it before committing. Pure bug fixes do **not** require a changeset.
    **Breaking changesets must carry their migration.** If the change removes or renames anything an author can write (a spec key, an export, a config field), the changeset body must state the FROM → TO mapping and the one-line fix — this text ships to consumers as `CHANGELOG.md` inside the npm package and is what an upgrading agent greps after the tombstone error. Removing an authorable spec key also requires a tombstone so the rejection itself carries the prescription — `retiredKey()` (`packages/spec/src/shared/retired-key.ts`) on a non-strict schema, or an entry in the relevant `UNKNOWN_KEY_GUIDANCE` / `*_RETIRED_KEY_GUIDANCE` map (see `object.zod.ts`, `ai/tool.zod.ts`) when the schema is `.strict()`. The changeset is one of fourteen surfaces a retirement touches — follow the `spec-property-retirement` skill (`.claude/skills/`) rather than reconstructing the kit, and note the two routes imply **opposite** liveness-ledger dispositions.
+   **A breaking changeset must also state its ADR-0087 disposition, in writing.** Add exactly one marker to the changeset body — `pnpm check:adr-0087-registration` enforces it, and the CI step is *Require an ADR-0087 disposition on a declared-breaking changeset*:
+   ```
+   <!-- adr-0087: registered SOME-MIGRATION-ID -->
+   <!-- adr-0087: not-required (unpublished) why -->
+   <!-- adr-0087: not-required (already-registered SOME-MIGRATION-ID) why -->
+   <!-- adr-0087: not-required (no-migration-prescription) why -->
+   ```
+   Why it is asked of you at all: the two ADR-0087 gates (`check:spec-changes`, `check:upgrade-guide`) pin ledger ↔ **artifact synchrony**, and the artifacts are a pure projection of the registry — so a retirement whose entry was **never written** leaves the two perfectly consistent and every gate in the repo green. PR #6048 removed `ctx.user.roles` that way and only a human comparing by eye caught it (#6011, backfilled by PR #6138). Ledger entries are the sole data source for `objectstack migrate meta`, `spec-changes.json` and the generated upgrade guide, and for a surface with **no spec schema** (`ctx.user` is only a runtime TS interface) there is no tombstone and no schema rejection either — the ledger entry is the *only* channel that reaches an upgrader. Measured: roughly **1 declared-breaking change in 7** needs an entry, so `not-required` is the ordinary answer and costs one line. Three of the four dispositions are re-verified mechanically on every run, and the fourth is refused when the changeset's own body carries a FROM → TO prescription — a changeset that ships migration instructions cannot also claim nobody must migrate (#6148).
 4. **Added or removed a `packages/spec` export? Run `pnpm --filter @objectstack/spec gen:api-surface` and commit the result.** The `TypeScript Type Check` job diffs spec's built export surface against `api-surface/` (one shard per entry point since #5837); a new export makes the snapshot stale and turns the job red. It reads the **built `dist` declarations**, so `OS_SKIP_DTS=1` — the flag you reach for to make local builds fast — skips exactly the artifact the gate inspects, and the check passes locally while failing in CI. Same shape for the other generated-artifact gates in that job (`check:docs`, `check:skill-refs`, `check:react-blocks`), which read `src/` and so do reproduce locally.
 5. Update `CHANGELOG.md` / `ROADMAP.md` if user-facing or architectural.
 6. **Delete temporary artifacts** — screenshots, traces, scratch logs, `.playwright-mcp/`, throwaway `tmp*.ts`, ad-hoc scripts. Repo must look identical to before, minus intended changes.

package.json

@@ -57,6 +57,7 @@
     "check:objectui-pin-fresh": "node scripts/check-objectui-pin-fresh.mjs --self-test && node scripts/check-objectui-pin-fresh.mjs",
     "check:prerelease-pins": "node scripts/check-prerelease-pin-watch.mjs --self-test && node scripts/check-prerelease-pin-watch.mjs",
     "check:empty-changeset": "node scripts/check-empty-changeset.mjs --self-test && node scripts/check-empty-changeset.mjs",
+    "check:adr-0087-registration": "node scripts/check-adr-0087-registration.mjs --self-test && node scripts/check-adr-0087-registration.mjs",
     "check:override-consistency": "node scripts/check-override-consistency.mjs --self-test && node scripts/check-override-consistency.mjs",
     "check:release-notes": "node scripts/check-release-notes.mjs",
     "check:release-body": "node scripts/release-github-releases.mjs --self-test",