@mulmoclaude/core 4.2.0 → 4.3.0

package/assets/helps/collection-skills.md

@@ -193,12 +193,23 @@ skipped, never crashes the host):
 
 Every field spec needs a `type` and a `label`. Extra keys by type:
 
-- **`datetime`** — no extra keys. Stored as a `YYYY-MM-DDTHH:MM` string and
-  edited with a native date+time picker. Use it (as `calendarField` /
-  `calendarEndField`) when an event has a real start/end clock — the calendar's
-  day view then draws each record as a proportional time block. For the common
-  "date column + separate time column" shape, keep `date` and point
-  `calendarTimeField` at the time string instead.
+- **`datetime`** — no extra keys. Stored as a `YYYY-MM-DDTHH:MM` string (seconds
+  optional) and edited with a native date+time picker. It is a **local wall
+  clock, not an instant**: no `Z`, no `+09:00` offset. `08:00` means eight in the
+  morning wherever the records are read, which is what a schedule means and what
+  the calendar can place. So a generated value must be FORMATTED, never
+  converted — `new Date(...).toISOString()` is wrong twice over: it appends the
+  `Z` (`putItems` writes the row but reports it in `lint`, a full `getItems`
+  listing warns about it, and a shared app's publish refuses it outright), and
+  it shifts the hours by the generating machine's offset, so a Tokyo 08:00 written on a US laptop reads back as the afternoon.
+  Build the string from the parts you already have (`${date}T${hh}:${mm}`). The
+  one `Z`-suffixed datetime that is legal is a shared app's server-stamped
+  field — nine fractional digits, written by the SERVER, never by you. Anything
+  else ending in `Z` (including `toISOString()`'s three digits) is linted.
+  Use it (as `calendarField` / `calendarEndField`) when an event has a real
+  start/end clock — the calendar's day view then draws each record as a
+  proportional time block. For the common "date column + separate time column"
+  shape, keep `date` and point `calendarTimeField` at the time string instead.
 - **`enum`** — `values: ["draft", "sent", "paid"]` (non-empty strings). Renders
   a `<select>`; stored as a plain string. Optional `default: "draft"` pre-fills
   a NEW record — the Add form opens on it, and a `putItems` row in `create` mode

package/assets/helps/error-recovery.md

@@ -1071,6 +1071,59 @@ Reading the refusal you got:
 Do NOT spawn the MCP bridge yourself. A hand-written JSON-RPC client fails
 invisibly and can leave a partially written collection that nothing reports.
 
+## `putItems` came back with a `lint` block — the rows were written, the values are the wrong shape
+
+### Symptoms
+
+- A `putItems` result carries `lint` beside `written`: `{ total, note, rows: [{ id, problem }] }`.
+- The problem text reads like a type complaint rather than a missing field —
+  `not a YYYY-MM-DDTHH:MM datetime`, `is not numeric`, `is not a real YYYY-MM-DD date`,
+  `is not a boolean`.
+- `rejected` is empty, and the records are on disk, so nothing looks broken yet.
+
+### Cause
+
+The write gate refuses only what makes a record unopenable: required fields,
+enum values, `primaryKey` = record id. The SHAPE of a value is checked by a
+second, report-only tier — refusing there would make a collection whose older
+rows predate the typed rules unwritable. So the row lands, and `lint` is the
+report.
+
+It is not cosmetic. Every later reader runs the strict tier: a full `getItems`
+listing (and a selective one that missed an id) returns the same finding as a
+`warning` beside the rows, and publishing a shared app REFUSES these rows
+outright. A batch generated by a script is wrong in every row the same way, so
+this is the moment it costs the least to fix.
+
+### Fix
+
+Correct the generator, then rewrite the rows (`mode: "upsert"`, same ids) and
+check that `lint` is gone. `lint.total` is the real count; `lint.rows` shows the
+first ten.
+
+The one that recurs is `datetime`. It stores a **local wall clock** —
+`YYYY-MM-DDTHH:MM`, seconds optional, **no `Z` and no offset** — because `08:00`
+in a schedule means eight in the morning wherever it is read, and that is what
+the calendar can place. So format the string, never convert it:
+
+```js
+// WRONG — appends `Z` and shifts the hours by the generating machine's offset
+{ startAt: new Date(`${day}T08:00`).toISOString() }   // "2026-08-17T15:00:00.000Z"
+
+// RIGHT — the clock you meant, written as the clock
+{ startAt: `${day}T08:00` }                            // "2026-08-17T08:00"
+```
+
+The conversion is the more expensive half: had the format passed, a Tokyo court's
+08:00 seeded on a US laptop would have published every slot seven hours out
+(eight, on the dates the other side of a DST change) with nothing to point at.
+The one `Z`-suffixed datetime the strict tier accepts is a shared app's
+server-stamped instant — nine fractional digits
+(`2026-08-15T01:45:54.605987654Z`), written by the server, not by you.
+`toISOString()`'s three digits are not that shape and are linted.
+
+Field-by-field stored forms: `schemaDocs` with `topic: "Field types"`.
+
 ## A tool your own instructions describe returns "No such tool available"
 
 ### Symptoms

package/dist/collection/registry/server/index.cjs

@@ -6,7 +6,7 @@ const require_relPath = require("../../../relPath-CTAjGdCL.cjs");
 const require_utils_fetch = require("../../../utils/fetch.cjs");
 const require_discovery = require("../../../discovery-Dl9nRa-m.cjs");
 const require_templatePath = require("../../../templatePath-27vUZowm.cjs");
-const require_server = require("../../../server-DZkN0GGh.cjs");
+const require_server = require("../../../server-CF1qllGV.cjs");
 const require_skill_bridge_index = require("../../../skill-bridge/index.cjs");
 const require_types = require("../../../types-CNqkLT4M.cjs");
 let node_fs = require("node:fs");

package/dist/collection/registry/server/index.js

@@ -4,7 +4,7 @@ import { n as toPosixRelPath } from "../../../relPath-DW8MC8VO.js";
 import { fetchWithTimeout } from "../../../utils/fetch.js";
 import { $ as collectionsRegistriesConfigPath, G as safeRecordId, ct as projectSkillsDir, mt as skillsStagingDir, nt as getWorkspaceRoot, ot as log, r as loadCollection, s as CollectionSchemaZ, t as acceptParsedSchema } from "../../../discovery-P64fgtuD.js";
 import { t as isSafeActionTemplatePath } from "../../../templatePath-k_WNbL_Q.js";
-import { v as ONE_SECOND_MS } from "../../../server-C-Ngd3g2.js";
+import { y as ONE_SECOND_MS } from "../../../server-BvXO49Vj.js";
 import { claudeSkillDir, dataSkillDir, mirrorSkillWrite } from "../../../skill-bridge/index.js";
 import { n as parseRegistryIndex, r as isRecord, t as OFFICIAL_REGISTRY_NAME } from "../../../types-BKVZrtyc.js";
 import { readFileSync } from "node:fs";

package/dist/collection/server/index.cjs

@@ -3,7 +3,7 @@ const require_root = require("../../root-rPH6FGDT.cjs");
 const require_itemId = require("../../itemId-CEQdK01z.cjs");
 const require_discovery = require("../../discovery-Dl9nRa-m.cjs");
 const require_templatePath = require("../../templatePath-27vUZowm.cjs");
-const require_server = require("../../server-DZkN0GGh.cjs");
+const require_server = require("../../server-CF1qllGV.cjs");
 exports.APP_MANIFEST_FILE = require_discovery.APP_MANIFEST_FILE;
 exports.BackendUnavailableError = require_discovery.BackendUnavailableError;
 exports.COLLECTION_ROOT_REQUIRED = require_discovery.COLLECTION_ROOT_REQUIRED;
@@ -14,6 +14,7 @@ exports.DEFAULT_QUERY_ROWS = require_discovery.DEFAULT_QUERY_ROWS;
 exports.MAX_CSV_ROWS = require_discovery.MAX_CSV_ROWS;
 exports.MAX_ITEMS_FILE_BYTES = require_server.MAX_ITEMS_FILE_BYTES;
 exports.MAX_PUT_ITEMS = require_server.MAX_PUT_ITEMS;
+exports.MAX_PUT_LINT = require_server.MAX_PUT_LINT;
 exports.MAX_QUERY_ROWS = require_discovery.MAX_QUERY_ROWS;
 exports.MAX_RECORD_ISSUES = require_server.MAX_RECORD_ISSUES;
 exports.MAX_SCHEMA_ISSUES = require_server.MAX_SCHEMA_ISSUES;

package/dist/collection/server/index.js

@@ -2,5 +2,5 @@ import { t as canonicalRoot } from "../../root-BMroU_mB.js";
 import { B as COMPUTED_TYPES } from "../../itemId-BY2YPBkT.js";
 import { A as readItem, B as SCHEMA_FILE, C as CollectionQueryZ, D as generateItemId, E as deleteItem, F as loadAppManifest, G as safeRecordId, H as itemFilePath, I as parseAppManifest, J as isBackendUnavailable, K as safeSlugName, L as sharedItemsPath, M as writeItem, N as APP_MANIFEST_FILE, O as isRegularFile, P as appManifestReason, Q as collectionChangePayload, R as pageFromFullRead, S as compileJsonlQuery, T as MAX_QUERY_ROWS, U as resolveDataDir, V as isContainedInRoot, W as resolveTemplatePath, Y as COLLECTION_ROOT_REQUIRED, Z as collectionChangeKey, _ as dedupeByRecordId, a as toDetail, at as localCollectionKey, d as storeFor, dt as setFirestoreAccessor, et as configureCollectionHost, ft as setSharedCollectionsSupport, g as decodeCsvRecordId, h as csvRowToItem, i as resolvePrimaryField, j as resolveCreateItemId, k as listItems, l as collectionWritable, lt as publishCollectionChange, n as discoverCollections, nt as getWorkspaceRoot, o as toSummary, ot as log, p as MAX_CSV_ROWS, pt as sharedCollectionChangePayload, q as BackendUnavailableError, r as loadCollection, rt as hostSupportsSharedCollections, s as CollectionSchemaZ, st as peekWorkspaceRoot, t as acceptParsedSchema, tt as firestoreHandle, u as readOnlyRefusal, ut as setCollectionChangePublisher, v as encodeCsvRecordId, w as DEFAULT_QUERY_ROWS, x as compileCsvQuery, y as normalizeCsvValue, z as projectItemFields } from "../../discovery-P64fgtuD.js";
 import { i as isSafeTemplatePath, n as isSafeCustomViewI18nPath, r as isSafeCustomViewPath, t as isSafeActionTemplatePath } from "../../templatePath-k_WNbL_Q.js";
-import { A as runCollectionQuery, C as firstMutateParamProblem, D as validateRecordObject, E as validateCollectionRecords, F as promptPathsFor, I as readCustomViewHtml, L as readCustomViewI18n, M as runQueryOverRows, N as buildActionSeedPrompt, O as compileRecordZ, P as buildCollectionActionSeedPrompt, R as readSkillTemplate, S as applyMutateAction, T as STORE_UNREADABLE, _ as successorId, a as makeManageCollectionTool, b as buildWorkspaceOntology, c as deleteCollectionRefusalMessage, d as daysInMonth, f as formatCivil, g as resolveEvery, h as parseCivil, i as MAX_UNSELECTIVE_ITEMS, j as enrichItems, k as recordFieldProblem, l as advanceTriggerDate, m as maybeSpawnSuccessor, n as MAX_PUT_ITEMS, o as deleteCustomView, p as isTriggerDue, r as MAX_SCHEMA_ISSUES, s as deleteCollection, t as MAX_ITEMS_FILE_BYTES, u as computeSuccessor, w as MAX_RECORD_ISSUES, x as schemaRelations, y as computeCollectionIcon } from "../../server-C-Ngd3g2.js";
-export { APP_MANIFEST_FILE, BackendUnavailableError, COLLECTION_ROOT_REQUIRED, COMPUTED_TYPES, CollectionQueryZ, CollectionSchemaZ, DEFAULT_QUERY_ROWS, MAX_CSV_ROWS, MAX_ITEMS_FILE_BYTES, MAX_PUT_ITEMS, MAX_QUERY_ROWS, MAX_RECORD_ISSUES, MAX_SCHEMA_ISSUES, MAX_UNSELECTIVE_ITEMS, SCHEMA_FILE, STORE_UNREADABLE, acceptParsedSchema, advanceTriggerDate, appManifestReason, applyMutateAction, buildActionSeedPrompt, buildCollectionActionSeedPrompt, buildWorkspaceOntology, canonicalRoot, collectionChangeKey, collectionChangePayload, collectionWritable, compileCsvQuery, compileJsonlQuery, compileRecordZ, computeCollectionIcon, computeSuccessor, configureCollectionHost, csvRowToItem, daysInMonth, decodeCsvRecordId, dedupeByRecordId, deleteCollection, deleteCollectionRefusalMessage, deleteCustomView, deleteItem, discoverCollections, encodeCsvRecordId, enrichItems, firestoreHandle, firstMutateParamProblem, formatCivil, generateItemId, getWorkspaceRoot, hostSupportsSharedCollections, isBackendUnavailable, isContainedInRoot, isRegularFile, isSafeActionTemplatePath, isSafeCustomViewI18nPath, isSafeCustomViewPath, isSafeTemplatePath, isTriggerDue, itemFilePath, listItems, loadAppManifest, loadCollection, localCollectionKey, log, makeManageCollectionTool, maybeSpawnSuccessor, normalizeCsvValue, pageFromFullRead, parseAppManifest, parseCivil, peekWorkspaceRoot, projectItemFields, promptPathsFor, publishCollectionChange, readCustomViewHtml, readCustomViewI18n, readItem, readOnlyRefusal, readSkillTemplate, recordFieldProblem, resolveCreateItemId, resolveDataDir, resolveEvery, resolvePrimaryField, resolveTemplatePath, runCollectionQuery, runQueryOverRows, safeRecordId, safeSlugName, schemaRelations, setCollectionChangePublisher, setFirestoreAccessor, setSharedCollectionsSupport, sharedCollectionChangePayload, sharedItemsPath, storeFor, successorId, toDetail, toSummary, validateCollectionRecords, validateRecordObject, writeItem };
+import { A as recordFieldProblem, C as applyMutateAction, D as validateCollectionRecords, E as STORE_UNREADABLE, F as buildCollectionActionSeedPrompt, I as promptPathsFor, L as readCustomViewHtml, M as enrichItems, N as runQueryOverRows, O as validateRecordObject, P as buildActionSeedPrompt, R as readCustomViewI18n, S as schemaRelations, T as MAX_RECORD_ISSUES, _ as resolveEvery, a as MAX_UNSELECTIVE_ITEMS, b as computeCollectionIcon, c as deleteCollection, d as computeSuccessor, f as daysInMonth, g as parseCivil, h as maybeSpawnSuccessor, i as MAX_SCHEMA_ISSUES, j as runCollectionQuery, k as compileRecordZ, l as deleteCollectionRefusalMessage, m as isTriggerDue, n as MAX_PUT_ITEMS, o as makeManageCollectionTool, p as formatCivil, r as MAX_PUT_LINT, s as deleteCustomView, t as MAX_ITEMS_FILE_BYTES, u as advanceTriggerDate, v as successorId, w as firstMutateParamProblem, x as buildWorkspaceOntology, z as readSkillTemplate } from "../../server-BvXO49Vj.js";
+export { APP_MANIFEST_FILE, BackendUnavailableError, COLLECTION_ROOT_REQUIRED, COMPUTED_TYPES, CollectionQueryZ, CollectionSchemaZ, DEFAULT_QUERY_ROWS, MAX_CSV_ROWS, MAX_ITEMS_FILE_BYTES, MAX_PUT_ITEMS, MAX_PUT_LINT, MAX_QUERY_ROWS, MAX_RECORD_ISSUES, MAX_SCHEMA_ISSUES, MAX_UNSELECTIVE_ITEMS, SCHEMA_FILE, STORE_UNREADABLE, acceptParsedSchema, advanceTriggerDate, appManifestReason, applyMutateAction, buildActionSeedPrompt, buildCollectionActionSeedPrompt, buildWorkspaceOntology, canonicalRoot, collectionChangeKey, collectionChangePayload, collectionWritable, compileCsvQuery, compileJsonlQuery, compileRecordZ, computeCollectionIcon, computeSuccessor, configureCollectionHost, csvRowToItem, daysInMonth, decodeCsvRecordId, dedupeByRecordId, deleteCollection, deleteCollectionRefusalMessage, deleteCustomView, deleteItem, discoverCollections, encodeCsvRecordId, enrichItems, firestoreHandle, firstMutateParamProblem, formatCivil, generateItemId, getWorkspaceRoot, hostSupportsSharedCollections, isBackendUnavailable, isContainedInRoot, isRegularFile, isSafeActionTemplatePath, isSafeCustomViewI18nPath, isSafeCustomViewPath, isSafeTemplatePath, isTriggerDue, itemFilePath, listItems, loadAppManifest, loadCollection, localCollectionKey, log, makeManageCollectionTool, maybeSpawnSuccessor, normalizeCsvValue, pageFromFullRead, parseAppManifest, parseCivil, peekWorkspaceRoot, projectItemFields, promptPathsFor, publishCollectionChange, readCustomViewHtml, readCustomViewI18n, readItem, readOnlyRefusal, readSkillTemplate, recordFieldProblem, resolveCreateItemId, resolveDataDir, resolveEvery, resolvePrimaryField, resolveTemplatePath, runCollectionQuery, runQueryOverRows, safeRecordId, safeSlugName, schemaRelations, setCollectionChangePublisher, setFirestoreAccessor, setSharedCollectionsSupport, sharedCollectionChangePayload, sharedItemsPath, storeFor, successorId, toDetail, toSummary, validateCollectionRecords, validateRecordObject, writeItem };

package/dist/collection/server/manageTool.d.ts

@@ -11,6 +11,11 @@ export declare const MAX_SCHEMA_ISSUES = 20;
  *  write that reported success would leave a half-filled collection nobody
  *  knows is half-filled. */
 export declare const MAX_PUT_ITEMS = 1000;
+/** Cap the strict-tier findings one putItems call reports back. The rows were
+ *  WRITTEN, so this is a report and not a refusal — and a generated batch is
+ *  wrong the same way in every row, which means ten samples plus the total say
+ *  everything a thousand would. */
+export declare const MAX_PUT_LINT = 10;
 /** Refuse an `itemsFile` larger than this, from `stat` and before any read.
  *  The row cap alone cannot bound the work: the file has to be read and parsed
  *  WHOLE before there are rows to count, so a huge blob is paid for in full
@@ -70,6 +75,25 @@ export type ManageCollectionDeps = DiscoveryOptions & {
      *  the sandbox is off — the prefix simply never appears. */
     sandboxWorkspacePath?: string | undefined;
 };
+/** One row the call did not take, or took and flagged: the id it was about and
+ *  what to fix. Exported alongside `PutItemsLint`, whose `rows` are these. */
+export interface RejectedRow {
+    id: string;
+    problem: string;
+}
+/** The `lint` block of a putItems result — the strict-tier findings on rows the
+ *  call WROTE. Exported because it is part of the tool's answer, not an
+ *  internal: a caller (or a test) that re-declares the shape inline is a copy
+ *  that stops matching the day the block gains a field. */
+export interface PutItemsLint {
+    /** Every flagged row, not just the shown ones — a capped `rows` must never be
+     *  readable as a total. */
+    total: number;
+    /** What the reader has to know to act on findings sitting beside `written`. */
+    note: string;
+    /** The first `MAX_PUT_LINT` findings. */
+    rows: RejectedRow[];
+}
 export declare function makeManageCollectionTool(deps?: ManageCollectionDeps): {
     definition: {
         name: string;

package/dist/collection-watchers/index.cjs

@@ -4,7 +4,7 @@ const require_itemId = require("../itemId-CEQdK01z.cjs");
 const require_promptSafety = require("../promptSafety-C0BWDsfT.cjs");
 require("../collection/index.cjs");
 const require_discovery = require("../discovery-Dl9nRa-m.cjs");
-const require_server = require("../server-DZkN0GGh.cjs");
+const require_server = require("../server-CF1qllGV.cjs");
 const require_notifier = require("../notifier-DZJj0sOh.cjs");
 let node_fs_promises = require("node:fs/promises");
 //#region src/collection-watchers/config.ts

package/dist/collection-watchers/index.js

@@ -3,7 +3,7 @@ import { O as isValidCollectionName, q as fieldText } from "../itemId-BY2YPBkT.j
 import { E as whenMatches, f as itemIsDone } from "../promptSafety-Bp780ezk.js";
 import "../collection/index.js";
 import { H as itemFilePath, Q as collectionChangePayload, d as storeFor, lt as publishCollectionChange, n as discoverCollections, pt as sharedCollectionChangePayload, r as loadCollection, st as peekWorkspaceRoot, tt as firestoreHandle } from "../discovery-P64fgtuD.js";
-import { m as maybeSpawnSuccessor, p as isTriggerDue } from "../server-C-Ngd3g2.js";
+import { h as maybeSpawnSuccessor, m as isTriggerDue } from "../server-BvXO49Vj.js";
 import { d as publish, m as updateForPlugin, n as clear, s as listAll } from "../notifier-DhnwR82U.js";
 import { access } from "node:fs/promises";
 //#region src/collection-watchers/config.ts

package/dist/feeds/server/index.cjs

@@ -5,7 +5,7 @@ const require_root = require("../../root-rPH6FGDT.cjs");
 const require_itemId = require("../../itemId-CEQdK01z.cjs");
 const require_project = require("../../project-C1ep9pvo.cjs");
 const require_discovery = require("../../discovery-Dl9nRa-m.cjs");
-const require_server = require("../../server-DZkN0GGh.cjs");
+const require_server = require("../../server-CF1qllGV.cjs");
 const require_feeds_paths = require("../paths.cjs");
 const require_ingestTypes = require("../../ingestTypes-CaX7_s-K.cjs");
 const require_notifier = require("../../notifier-DZJj0sOh.cjs");

package/dist/feeds/server/index.js

@@ -3,7 +3,7 @@ import { t as canonicalRoot } from "../../root-BMroU_mB.js";
 import { H as INGEST_KINDS, V as FEED_SCHEDULES } from "../../itemId-BY2YPBkT.js";
 import { t as mergeIntoExisting } from "../../project-B0j_Vf_Q.js";
 import { K as safeSlugName, U as resolveDataDir, _t as createHostSlot, d as storeFor, gt as createForwardingLogger, k as listItems, n as discoverCollections } from "../../discovery-P64fgtuD.js";
-import { F as promptPathsFor, P as buildCollectionActionSeedPrompt, R as readSkillTemplate } from "../../server-C-Ngd3g2.js";
+import { F as buildCollectionActionSeedPrompt, I as promptPathsFor, z as readSkillTemplate } from "../../server-BvXO49Vj.js";
 import { FEEDS_DIR, feedDir, feedStatePath, feedsRoot, ingestStateDir, ingestStatePath } from "../paths.js";
 import { n as DEFAULT_FEED_MAX_ITEMS, r as isFeedSchedule, t as AGENT_INGEST_KIND } from "../../ingestTypes-CIy1lHjf.js";
 import { d as publish, n as clear } from "../../notifier-DhnwR82U.js";

package/dist/{server-C-Ngd3g2.js → server-BvXO49Vj.js}

@@ -1566,6 +1566,22 @@ var MAX_SCHEMA_ISSUES = 20;
 *  write that reported success would leave a half-filled collection nobody
 *  knows is half-filled. */
 var MAX_PUT_ITEMS = 1e3;
+/** Cap the strict-tier findings one putItems call reports back. The rows were
+*  WRITTEN, so this is a report and not a refusal — and a generated batch is
+*  wrong the same way in every row, which means ten samples plus the total say
+*  everything a thousand would. */
+var MAX_PUT_LINT = 10;
+/** What the `lint` block SAYS, once per call. A bare list of findings on rows the
+*  call also reports as `written` reads as noise; the reason to act is that the
+*  next reader of these rows is stricter than the write was.
+*
+*  Like everything else the tool says about validation, this describes the tool
+*  with its gate ON. Under `ablateValidation` nothing is checked — `rejected`
+*  stays empty too — and the caveat is deliberately NOT written into the
+*  agent-facing text: ablation exists to measure what a model does without the
+*  guardrails, and a model told "unless ablation is on" is no longer without
+*  them (CodeRabbit on #2925). */
+var PUT_LINT_NOTE = "These rows WERE written. The write gate refuses what would make a record unopenable — a missing required field, a value outside an enum, a mismatched primaryKey, a computed key, a colliding id under `create` — but it does not check the SHAPE of a value, so a wrong-shaped one is reported here rather than rejected. `getItems` surfaces the same finding as a `warning` (on a full listing, or when a requested id is missing), and publishing a shared app REFUSES the row outright. Fix the generator and rewrite them before writing the rest of the set.";
 /** Refuse an `itemsFile` larger than this, from `stat` and before any read.
 *  The row cap alone cannot bound the work: the file has to be read and parsed
 *  WHOLE before there are rows to count, so a huge blob is paid for in full
@@ -1734,6 +1750,26 @@ async function mergeWithExisting(collection, store, record, itemId) {
 		...record
 	};
 }
+/** The strict-tier finding on a row the write gate LET THROUGH, or nothing.
+*
+*  The gate stays exactly where it was — `lint-not-lock` is why the strict tier
+*  exists (see `../core/recordZ`), and turning these into rejections would make
+*  a collection whose legacy rows predate the typed rules unwritable. What was
+*  missing is that the row's author never HEARD about it: `getItems` reports the
+*  same finding and publishing a shared app refuses exactly these rows, so the
+*  one surface that stayed silent was the one that could still fix the generator
+*  before it wrote the other 719 rows (mulmoterminal#1763).
+*
+*  Run against the record as WRITTEN (`toWrite`, post-merge, post-defaults) —
+*  what a later read will lint is what landed, not what the caller sent. */
+function lintOf(record, itemId, schema, deps) {
+	if (deps.ablateValidation) return {};
+	const problem = validateRecordObject(record, itemId, schema, "strict");
+	return problem ? { lint: {
+		id: defangForPrompt(itemId),
+		problem: defangForPrompt(problem)
+	} } : {};
+}
 async function putOneItem(collection, store, write, record, mode, deps) {
 	const { schema } = collection;
 	const created = mode === "create" ? {
@@ -1759,7 +1795,10 @@ async function putOneItem(collection, store, write, record, mode, deps) {
 		if (invalid) return reject(itemId, invalid);
 	}
 	const result = await write(itemId, toWrite, { refuseOverwrite: mode === "create" });
-	if (result.kind === "ok") return { written: result.itemId };
+	if (result.kind === "ok") return {
+		written: result.itemId,
+		...lintOf(toWrite, result.itemId, schema, deps)
+	};
 	if (result.kind === "invalid-id") return reject(itemId, `'${itemId}' is not a valid record id (letters/digits at the ends; -, _, or . inside; no '..' or path characters)`);
 	if (result.kind === "conflict") return reject(itemId, `'${itemId}' already exists — mode "create" refuses overwrite; use "upsert" to update it`);
 	return reject(itemId, "write refused: the collection's data dir escapes the workspace");
@@ -1932,23 +1971,51 @@ async function resolvePutRows(args, deps) {
 	if (rows.length > 1e3) return `manageCollection: refused — ${rows.length} rows is over the putItems limit of ${MAX_PUT_ITEMS}. Nothing was written; split them across several calls.`;
 	return rows;
 }
-async function handlePutItems(collection, args, deps) {
-	const store = storeFor(collection, { workspaceRoot: deps.workspaceRoot });
-	const { write } = store;
-	if (!write) return `manageCollection: ${readOnlyRefusal(collection.slug)} (its records are the rows of '${collection.schema.dataSource?.path}'; edit that file to change the data).`;
-	const rows = await resolvePutRows(args, deps);
-	if (typeof rows === "string") return rows;
+/** The `lint` block of a putItems result, or nothing when every written row is
+*  clean — absent rather than empty, so a clean call reads exactly as it always
+*  did and the key's presence is itself the signal.
+*
+*  `total` counts every flagged row and `rows` shows the first
+*  `MAX_PUT_LINT` of them, so a capped report can never be read as a total. */
+function lintReport(lint) {
+	if (lint.length === 0) return {};
+	return { lint: {
+		total: lint.length,
+		note: PUT_LINT_NOTE,
+		rows: lint.slice(0, 10)
+	} };
+}
+/** Write the batch row by row, sorting each outcome into its own list. One row
+*  at a time on purpose: a rejection is per row, so a bad row must not take the
+*  batch with it. */
+async function putEachRow(collection, store, write, rows, mode, deps) {
 	const written = [];
 	const rejected = [];
+	const lint = [];
 	for (const record of rows) {
-		const outcome = await putOneItem(collection, store, write, record, args.mode, deps);
+		const outcome = await putOneItem(collection, store, write, record, mode, deps);
 		if (outcome.written) written.push(outcome.written);
 		if (outcome.rejected) rejected.push(outcome.rejected);
+		if (outcome.lint) lint.push(outcome.lint);
 	}
+	return {
+		written,
+		rejected,
+		lint
+	};
+}
+async function handlePutItems(collection, args, deps) {
+	const store = storeFor(collection, { workspaceRoot: deps.workspaceRoot });
+	const { write } = store;
+	if (!write) return `manageCollection: ${readOnlyRefusal(collection.slug)} (its records are the rows of '${collection.schema.dataSource?.path}'; edit that file to change the data).`;
+	const rows = await resolvePutRows(args, deps);
+	if (typeof rows === "string") return rows;
+	const { written, rejected, lint } = await putEachRow(collection, store, write, rows, args.mode, deps);
 	return JSON.stringify({
 		collection: collection.slug,
 		written,
-		rejected
+		rejected,
+		...lintReport(lint)
 	});
 }
 /** Delete records by id, through the store — so it works on any writable
@@ -2158,7 +2225,7 @@ async function handlePutSchema(slug, schemaArg, deps) {
 		written: true
 	});
 }
-var MANAGE_COLLECTION_PROMPT = "Use `manageCollection` instead of raw Read/Write/Edit when working with a collection's records OR its schema (raw file I/O stays available as the escape hatch). Before authoring or changing a collection's `schema.json`, call `schemaDocs` to load the field/DSL reference — the default reply is the core authoring guide plus a table of contents; fetch advanced sections (actions, bells, calendar/kanban views, dataSource, storage) by passing their heading as `topic` rather than dumping `topic: \"all\"`. Then read with `getSchema` and write with `putSchema` — `putSchema` validates the whole schema before writing and returns actionable errors instead of silently failing discovery's validation. `getItems` is the only way to see computed values — `derived` fields (e.g. a portfolio's value), `toggle` projections, and `embed` records are host-computed and never present in the stored JSON files. On large collections pass `ids` and/or `fields` to keep the result small. For a question that spans collections (\"which clients have unpaid invoices?\"), start with `getOntology`: it lists every collection with its primaryKey, record count, and outbound `ref`/`embed` relations, so you know which collections to join before reading any records. `putItems` validates every row against the schema before writing (required fields, enum values, primaryKey = record id) and returns `{ written, rejected }`; fix each rejected row using its `problem` text and retry just those rows. Never include computed fields in a row you write. When the rows come from a script rather than from you (a generated schedule, an imported set, anything past a few dozen records), write them to a JSON file UNDER THE WORKSPACE and pass its absolute path as `itemsFile` instead of `items` — the host reads the file, so the rows never pass through your context. Do NOT hand-transcribe a generated file into `items`, and never drive the collection by spawning the MCP bridge yourself. To update a few fields of an existing record, use `mode: \"merge\"` with a partial row ({ id, <changed fields> }) — the default upsert replaces the WHOLE record, so a partial upsert would silently erase every optional field it omits. `deleteItems` removes records by id and returns `{ deleted, rejected }`; an id that doesn't exist comes back rejected rather than counted as deleted, so check `rejected` before reporting a deletion as done. Answer aggregation questions (counts, sums, averages, group-bys) with `queryItems` on ANY collection — on a dataSource (CSV) collection it scans the whole file (getItems is row-capped, so aggregates computed from its output can be silently wrong on large files); on a file-backed collection it aggregates the enriched records, so computed fields (derived/rollup/toggle) are queryable columns.";
+var MANAGE_COLLECTION_PROMPT = "Use `manageCollection` instead of raw Read/Write/Edit when working with a collection's records OR its schema (raw file I/O stays available as the escape hatch). Before authoring or changing a collection's `schema.json`, call `schemaDocs` to load the field/DSL reference — the default reply is the core authoring guide plus a table of contents; fetch advanced sections (actions, bells, calendar/kanban views, dataSource, storage) by passing their heading as `topic` rather than dumping `topic: \"all\"`. Then read with `getSchema` and write with `putSchema` — `putSchema` validates the whole schema before writing and returns actionable errors instead of silently failing discovery's validation. `getItems` is the only way to see computed values — `derived` fields (e.g. a portfolio's value), `toggle` projections, and `embed` records are host-computed and never present in the stored JSON files. On large collections pass `ids` and/or `fields` to keep the result small. For a question that spans collections (\"which clients have unpaid invoices?\"), start with `getOntology`: it lists every collection with its primaryKey, record count, and outbound `ref`/`embed` relations, so you know which collections to join before reading any records. `putItems` GATES every row on what would make the record unopenable — required fields, enum values, primaryKey = record id — and returns `{ written, rejected }`; fix each rejected row using its `problem` text and retry just those rows. Never include computed fields in a row you write. That gate does NOT check the SHAPE of a value: a `datetime` written as an instant (`2026-08-17T15:00:00.000Z`, what `toISOString()` produces) rather than as a wall clock, a `number` holding something that is not a number at all, a `date` that is not a real day are all WRITTEN, and come back in a `lint` block beside `written`. Read it — a full `getItems` listing warns about the same rows and publishing a shared app REFUSES them, so a silent `lint` is the only proof the values are right. Before generating a large set, read the exact stored form of each type in the `Field types` section of `schemaDocs` (`topic: \"Field types\"` fetches just that one), then write ONE batch and check that `lint` is absent. `datetime` in particular is a local wall clock (`YYYY-MM-DDTHH:MM`, no timezone suffix), so `new Date(...).toISOString()` is wrong twice over — the suffix, and the hours the conversion moved. The one `Z`-suffixed datetime the strict tier accepts is a shared app's server-stamped instant, written by the SERVER with nine fractional digits; you never produce that value. When the rows come from a script rather than from you (a generated schedule, an imported set, anything past a few dozen records), write them to a JSON file UNDER THE WORKSPACE and pass its absolute path as `itemsFile` instead of `items` — the host reads the file, so the rows never pass through your context. Do NOT hand-transcribe a generated file into `items`, and never drive the collection by spawning the MCP bridge yourself. To update a few fields of an existing record, use `mode: \"merge\"` with a partial row ({ id, <changed fields> }) — the default upsert replaces the WHOLE record, so a partial upsert would silently erase every optional field it omits. `deleteItems` removes records by id and returns `{ deleted, rejected }`; an id that doesn't exist comes back rejected rather than counted as deleted, so check `rejected` before reporting a deletion as done. Answer aggregation questions (counts, sums, averages, group-bys) with `queryItems` on ANY collection — on a dataSource (CSV) collection it scans the whole file (getItems is row-capped, so aggregates computed from its output can be silently wrong on large files); on a file-backed collection it aggregates the enriched records, so computed fields (derived/rollup/toggle) are queryable columns.";
 /** Validate getItems' optional `ids`/`fields` args, then delegate. */
 async function dispatchGetItems(collection, args, deps) {
 	const ids = optionalStringArray(args.ids, "ids");
@@ -2214,7 +2281,7 @@ async function dispatchManageCollection(deps, args) {
 }
 var MANAGE_COLLECTION_DEFINITION = {
 	name: "manageCollection",
-	description: "Read and write a schema-driven collection through the host — both its records and its structure. getItems returns records WITH computed values (derived formulas, toggles, embeds) the stored JSON files don't contain; putItems validates each row against the schema before writing; deleteItems removes records by id. getOntology maps the whole workspace: every collection with its record count and outbound ref/embed relations — call it first for cross-collection questions. schemaDocs returns the collection-authoring reference — the core guide plus a table of contents by default; pass `topic` for a specific section. getSchema/putSchema read and validate-then-write the collection's schema.json. Prefer it over raw file I/O on collections.",
+	description: "Read and write a schema-driven collection through the host — both its records and its structure. getItems returns records WITH computed values (derived formulas, toggles, embeds) the stored JSON files don't contain; putItems gates each row on required/enum/primaryKey before writing and lints the value shapes it does not gate; deleteItems removes records by id. getOntology maps the whole workspace: every collection with its record count and outbound ref/embed relations — call it first for cross-collection questions. schemaDocs returns the collection-authoring reference — the core guide plus a table of contents by default; pass `topic` for a specific section. getSchema/putSchema read and validate-then-write the collection's schema.json. Prefer it over raw file I/O on collections.",
 	inputSchema: {
 		type: "object",
 		properties: {
@@ -2289,6 +2356,6 @@ function makeManageCollectionTool(deps = {}) {
 	};
 }
 //#endregion
-export { runCollectionQuery as A, firstMutateParamProblem as C, validateRecordObject as D, validateCollectionRecords as E, promptPathsFor as F, readCustomViewHtml as I, readCustomViewI18n as L, runQueryOverRows as M, buildActionSeedPrompt as N, compileRecordZ as O, buildCollectionActionSeedPrompt as P, readSkillTemplate as R, applyMutateAction as S, STORE_UNREADABLE as T, successorId as _, makeManageCollectionTool as a, buildWorkspaceOntology as b, deleteCollectionRefusalMessage as c, daysInMonth as d, formatCivil as f, resolveEvery as g, parseCivil as h, MAX_UNSELECTIVE_ITEMS as i, enrichItems as j, recordFieldProblem as k, advanceTriggerDate as l, maybeSpawnSuccessor as m, MAX_PUT_ITEMS as n, deleteCustomView as o, isTriggerDue as p, MAX_SCHEMA_ISSUES as r, deleteCollection as s, MAX_ITEMS_FILE_BYTES as t, computeSuccessor as u, ONE_SECOND_MS as v, MAX_RECORD_ISSUES as w, schemaRelations as x, computeCollectionIcon as y };
+export { recordFieldProblem as A, applyMutateAction as C, validateCollectionRecords as D, STORE_UNREADABLE as E, buildCollectionActionSeedPrompt as F, promptPathsFor as I, readCustomViewHtml as L, enrichItems as M, runQueryOverRows as N, validateRecordObject as O, buildActionSeedPrompt as P, readCustomViewI18n as R, schemaRelations as S, MAX_RECORD_ISSUES as T, resolveEvery as _, MAX_UNSELECTIVE_ITEMS as a, computeCollectionIcon as b, deleteCollection as c, computeSuccessor as d, daysInMonth as f, parseCivil as g, maybeSpawnSuccessor as h, MAX_SCHEMA_ISSUES as i, runCollectionQuery as j, compileRecordZ as k, deleteCollectionRefusalMessage as l, isTriggerDue as m, MAX_PUT_ITEMS as n, makeManageCollectionTool as o, formatCivil as p, MAX_PUT_LINT as r, deleteCustomView as s, MAX_ITEMS_FILE_BYTES as t, advanceTriggerDate as u, successorId as v, firstMutateParamProblem as w, buildWorkspaceOntology as x, ONE_SECOND_MS as y, readSkillTemplate as z };
 
-//# sourceMappingURL=server-C-Ngd3g2.js.map
+//# sourceMappingURL=server-BvXO49Vj.js.map