docs: three corrections to yesterday's ruling, found by re-reading the sources

**The correction bar overstated its own evidence.** It said "four times before
1.0 the loader accepted a form that could not work," then cited #342 and #335.
Only those two made metadata stop LOADING. The other two applications of the
doctrine — the 0.19.1 `@min` clamp and the 0.21.6 `like` case-sensitivity fix —
corrected OUTPUT and needed no exception at all, since RELEASING.md's
"output change = pure bugfix" row already covers them. Writing the count as
four made the exception look four times as load-bearing as it is, in the one
document where an adopter goes to find out what we are allowed to do to them.
Now split explicitly: two output, two load.

**The `agent-docs` migration instruction was invented.** I wrote "import from
the package root instead" from a summary rather than the changelog. The actual
entry says: delete the import, the live surface is
`@metaobjectsdev/sdk/agent-context`. A migration guide that names the wrong
replacement is worse than one that omits the entry, because the reader stops
looking.

**G3a's evidence is assembled** rather than left as an instruction to go and
find it. Three checkable facts: the registry carries zero `experimental`
markers; every 1.1 roadmap item is additive (FR-024 and FR-034 add reserved
vocabulary, FR-026/028/030 add no vocabulary at all); and all three open issues
are CLI surfaces over vocabulary that already exists. The mark stays 🔶 because
declaring the scope closed is the maintainer's call, not the evidence-gatherer's
— and the claim to declare is the weaker, true one: not that the vocabulary is
complete, but that completing it later adds and never changes or removes.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_01At3v6M6uqECZ2Sb5eUv6YY

Author: dmealing

docs/1.0-readiness.md

@@ -207,11 +207,27 @@ Status legend: ✅ done · 🔶 in progress · ⬜ not started · ❓ **[RATIFY]
   **What replaces it:**
 
   - 🔶 **G3a — the declared scope is covered, and no known gap needs NEW vocabulary.**
-    State what Metamodel 1.0 covers and show every outstanding item is additive. Half
-    ratified already: A2 puts FR-024's declared-API tier explicitly outside 1.0, and
-    `spec/roadmap.md` charters FR-037 R3/R4/R5 and the FR-038 stub generator as additive
-    1.1 work — **no breaking change is currently chartered**. Turn that into a positive
-    scope statement rather than a list of deferrals.
+    State what Metamodel 1.0 covers and show every outstanding item is additive. **The
+    evidence is assembled (2026-09-06); what remains is the maintainer declaring it.**
+
+    1. **Nothing in the registry is experimental.** `expected-registry.json` carries zero
+       `experimental` markers, and `docs/compatibility-policy.md` already excludes
+       "anything explicitly marked experimental or reserved and not yet in the registry"
+       from the covered set — so the covered surface IS the manifest at the cut commit.
+    2. **Every outstanding roadmap item is ADDITIVE.** 1.1 holds FR-024's declared-API tier
+       (`api.*`/`operation.*`/`binding.*` — reserved, unregistered, additive by
+       construction: A2), FR-034's ecosystem tier (same shape), FR-026's view-render polish
+       (TS web only), FR-028's object↔JSON parity + `meta export`, and FR-030's serializer
+       SPI. FR-037 R3/R4/R5 and the FR-038 stub generator are explicitly additive
+       (`spec/roadmap.md`). **No breaking change is chartered anywhere.**
+    3. **No open issue needs new vocabulary.** The three open at the cut — `meta why`
+       (#306), enforcing the already-shipped `@deprecated` (#305), and `meta fmt` (#304) —
+       are CLI surfaces and enforcement over vocabulary that already exists.
+
+    The claim to declare is therefore not "the vocabulary is complete" but the weaker and
+    true one: **completing it later adds, and never changes or removes.** That is what
+    ADR-0035 §1 already promises by excluding reserved-but-unregistered vocabulary from the
+    compat surface; G3a is where the project states it has checked, rather than assumed it.
   - ✅ **G3b — the 1.0 promise is livable: the correction bar is ratified.** DONE
     2026-09-06. `docs/compatibility-policy.md` now carries *"Correcting input we wrongly
     accepted"* — the three-part bar under which a form that never had a valid meaning may

docs/compatibility-policy.md

@@ -90,12 +90,16 @@ Everything above answers "what happens when we *change* the contract." This sect
 answers a different question: **what happens when we find that the toolchain accepted
 something the contract never allowed, and the fix makes it stop loading?**
 
-That case is real and it recurs. Four times before 1.0 the loader accepted a form that
-could not work — an index declaring both `@fields` and `@expr`, where one half was
-silently discarded ([#342](https://github.com/metaobjectsdev/metaobjects/issues/342));
+That case is real and it recurs. The doctrine behind it — *a correction of previously-wrong
+acceptance is a bug fix, not a contract change* — has been applied four times before 1.0.
+**Two of those changed only OUTPUT** and are ordinary bug fixes needing no exception: the
+`0.19.1` `@min` clamp (an authored `@min: 0` was silently discarded) and the `0.21.6` `like`
+case-sensitivity fix. **Two made previously-LOADING metadata stop loading**, and those are
+what this section governs: an index declaring both `@fields` and `@expr`, where one half was
+silently thrown away ([#342](https://github.com/metaobjectsdev/metaobjects/issues/342)), and
 `@filterable` on an array field, which emitted SQL that cannot execute
-([#335](https://github.com/metaobjectsdev/metaobjects/issues/335)). Refusing those is
-not a new rule. It is the documented rule finally being enforced.
+([#335](https://github.com/metaobjectsdev/metaobjects/issues/335)). Refusing those two is not
+a new rule. It is the documented rule finally being enforced.
 
 Under ADR-0023 the registry is strict and sealed, so there is no deprecation shim: a
 refusal takes effect on the release that ships it. That makes it important to say

docs/features/migrations/0.x-to-1.0.md

@@ -227,8 +227,12 @@ a table, a view and a stored procedure, told apart only by a sibling `kind`.
 
 **10E. The `@metaobjectsdev/sdk/agent-docs` subpath is removed**
 
-Import from the package root instead. The subpath shipped a blob that the new `agent/` docs
-surface replaces.
+Breaking only for code importing `AGENT_DOCS_BODY` from it. **Delete the import; the live
+surface is `@metaobjectsdev/sdk/agent-context`.** It was the pre-agent-context single-blob
+reference — `meta init` had already stopped scaffolding it, but it stayed exported while its
+prose rotted, teaching six things the loader rejects. It was deleted rather than corrected,
+because keeping it means two prompt surfaces held in agreement forever, one of which nothing
+assembles, tests or scaffolds.
 
 **10F. Python `GenConfig.column_naming` raises `TypeError`**