wp-creation-app.js:1144 said "Subject and WP Type are required" in an alert()
on a form ten cards deep, naming nothing, focusing nothing. The creator
carried 40 native call sites in all (43 at the wave 0 count; three had
already left with D5 and T5.8's wizard work).
Inline validation, the T5.8 wizard pattern applied to the creator:
- WP_REQUIRED is one table: field id, owning section, label. The error box,
aria-describedby, aria-invalid and the role=alert announcement all follow
from a row. The conditional IFF rule folded in beside them.
- Submit marks every failing field, marks the rail entry of each section
holding one (a "!" chip - a character, not only a colour), switches to the
section of the FIRST error, scrolls to and focuses the field, and announces
the failure through the role=alert toast.
One modal replaced confirm() and prompt(): promise-based wpConfirmDialog()/
wpPromptDialog() with an optional input whose validation renders AT the input
(a bad answer keeps the dialog open and says why - no round-trip through a
second dialog). Escape cancels; callers read like the natives they replaced,
awaited. Pure notifications became role-differentiated toasts. The modal
validation errors for the hold log and the QA rejection render inline in
their own modals.
The A1 path: confirmEarlyRelease() keeps its name and contract - truthy means
proceed with the reason recorded - and became async; every caller awaits it
(status control, hold release, urgent override, save).
App-wide native dialog count, recorded per the done-when: the probe prints it
against the wave 0 baseline of 79 and asserts the creator contributes 0. The
probe also replaces the natives with throwing stubs for the whole run, so any
path that still reached one would fail loudly.
hold_check re-pointed, not relaxed: three flows it drove through native
stubs now drive the modal - same propositions (the release-ready offer, the
named-constraints override prompt, the hard block), new surface.
Verification (each probe run alone): NEW tests/creator_dialogs_check.py
20/20. Regressions: hold_check 50/50 (re-pointed), warning_check 17/17,
qa_gate_check 40/40, triage_check 16/16, files_check 36/36, frame_check
39/39, generalinfo_check 49/49, form_structure_check 50/51 (the standing F6
height check - see the wave exit).
Items: S1 (creator half)
Co-Authored-By: Claude Fable 5 <noreply@anthropic.com>
