toga-ai 1.0.86 → 1.0.87

package/knowledge/1.0/apps/webhook/features/netsuite-clickup-opportunity-sync.md

@@ -57,9 +57,11 @@ Its output depends entirely on the trigger **context** (because it reads field v
   changed field on `context.newRecord` → `tranid`/entity/title read empty → 2493 falls back to
   `name = "OPP " + record.id` with empty descriptive/custom fields → the degenerate stub.
 
-The degenerate trigger is **scheduled script 2412 "Update Opp Status / % to Close / Presales Lead"**
+The degenerate trigger was **scheduled script 2412 "Update Opp Status / % to Close / Presales Lead"**
 (deployment `customdeploy_update_opp_status_morning`, status SCHEDULED). It auto-assigns the Presales
-Lead via `submitFields` (NetSuite context `SCH`, user `-System-`), which fires 2493 in `xedit` → stub.
+Lead (NetSuite context `SCH`, user `-System-`) — originally via `submitFields`, which fired 2493 in
+`xedit` → stub. **Fixed 2026-06-15:** 2412 now does a full `record.load()`/`setValue()`/`save()`, so
+the assignment fires 2493 in `edit` (full-record) context → proper name (see Gotchas → Fix).
 
 ## Data model
 - ClickUp Opportunities list `901111987449`; custom-item id `1009`.
@@ -85,17 +87,25 @@ None — single internal Agilant/True integration.
   June 1 edit added the Presales-Lead auto-assignment. **Precise failure condition:** only the
   *empty→set* transition fires it — a scheduled set to an already-present lead is a no-op (confirmed
   on opp 5483750, where a 07:02 `SCH` set produced no stub because the lead was already present).
-- **Fix options (NetSuite-side, in script 2493):**
-  - *Preferred* — `record.load({type, id: context.newRecord.id})` (or `search.lookupFields` for
-    `tranid`/entity/title) before building `name`. Works for **every** trigger context; produces
-    correct names instead of stubs-or-nothing.
-  - *Alternative* — skip on `xedit` (`if (context.type === context.UserEventType.XEDIT) return;`).
-    Suppresses the stubs but **also drops** legitimate task creation from inline-grid edits, mass
-    updates, and any other `submitFields` automation. Does **not** affect full-form saves, creates,
-    or REST/SOAP edits (incl. the touch tool, which is `RWS`/full-edit). Trades a visible bug for a
-    silent one — only acceptable if presales leads are never set inline/in bulk.
-  - *TOGA backstop (optional)* — in `createTask()`, if `name` is empty or matches `^OPP \d+$`, look
-    the opp up by internal id and rebuild the name before POSTing.
+- **Fix — applied in NetSuite prod 2026-06-15 (deployed; pending verification on next scheduled
+  run).** In scheduled script **2412**, the presales-lead assignment was changed from
+  `record.submitFields({type, id, values:{custbody_ctc_solution_architect: employeeId}, ...})` to a
+  full `record.load()` → `setValue('custbody_ctc_solution_architect', employeeId)` →
+  `record.save({ignoreMandatoryFields:true})`. A full `save()` fires 2493 with `context.type='edit'`
+  and the **complete** `newRecord`, so it reads `tranid`/entity/title and builds a proper name; the
+  empty→set guard still works (`oldRecord` reflects the pre-save state). **Verify** by watching the
+  next `SCH` empty→set run for a properly-named task (not `OPP <id>`) — not yet observed against a
+  live scheduled run as of the fix.
+  - **Residual exposure:** 2493 still reads off `context.newRecord`, so *other* `xedit`-context
+    presales-lead sets (inline-grid edit, Mass Update, any other script's `submitFields`) would still
+    stub. The 2412 fix closes the actual regression, not the latent 2493 hole. Belt-and-suspenders =
+    add a `record.load()`/`lookupFields` fallback **inside 2493** too (not done).
+  - **Watch-outs from the full save:** governance per opp rose ~2.5× (`load` ~5 + `save` ~20 units vs
+    `submitFields` ~10) — guard batch runs against the governance ceiling (yield/reschedule); and a
+    full `save()` re-fires **all** `edit`-context UserEvents/workflows on the opp (Quosal, LSA, etc.),
+    not just 2493.
+  - **TOGA backstop (optional, not done):** in `createTask()`, if `name` is empty or matches
+    `^OPP \d+$`, look the opp up by internal id and rebuild the name before POSTing.
 - **No inbound logging.** `Webhook_NetSuite::post()` has a `// Log payload to file` comment but does
   **nothing** — the raw NetSuite payload is persisted nowhere, which is why this was hard to trace.
   Adding real request logging here would make recurrences diagnosable immediately.
@@ -118,9 +128,13 @@ It blanks then restores `custbody_ctc_solution_architect` via REST (`RWS`/full-e
 with the full record → proper `<tranid> — <custNum> <customer> — <title>` task ~1–2s later.
 **Leaves the original stub** — delete the `OPP <id>` task in ClickUp afterward.
 Done 2026-06-15 for opps 7147699/7147698/7147472/6382879 → 4 proper tasks created; the 4 stubs
-(task ids `868k0tr41`, `868k0tr16`, `868k0rwtp`, `868k0h44m`) were left for deletion.
+(task ids `868k0tr41`, `868k0tr16`, `868k0rwtp`, `868k0h44m`) were archived in ClickUp afterward.
 
 ## Change history
+- 2026-06-15 — **Fix deployed in NetSuite prod:** scheduled script 2412's presales-lead assignment
+  switched from `submitFields` to full `load`/`setValue`/`save` → fires 2493 in `edit` (full-record)
+  context → proper names. Residual: 2493 still latently `xedit`-vulnerable to other triggers; watch
+  governance (~2.5×) and other UserEvents re-firing on the full save. Pending live-run verification. (dfranks)
 - 2026-06-15 — Identified the exact NetSuite culprits (scheduled **2412** `submitFields`→`xedit`→
   UserEvent **2493**), regression date (script 2412 modified 2026-06-01; no stubs before Jun 12),
   and `systemnote`/`SCH` confirmation; added remediation-via-touch procedure and fix-option

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "toga-ai",
-  "version": "1.0.86",
+  "version": "1.0.87",
   "description": "TOGA Technology Team Claude Knowledge System — shared AI coding harness with skills, knowledge base CLI, and project installer for Claude Code.",
   "keywords": [
     "claude",