feat(#2646): surface unresolved deferred-items.md at milestone close (#2983)

* feat(#2646): surface unresolved deferred-items.md at milestone close

auditOpenArtifacts scanned eight categories; deferred-items.md was not
among them. #2287 made that file readable at the PHASE boundary
(audit-uat, /gsd-progress check 7), but one boundary up it stayed
invisible — and phase directories archive to milestones/vX.Y-phases/ by
default (#1871), so an out-of-scope discovery a phase agent correctly
recorded rather than fixed left the live tree at milestone close having
never reached the [R]/[A]/[C] prompt that exists to catch exactly this.

Adds deferred_items as a ninth scanner plus its count, its items entry
and its report section. The workflow needed no change: complete-milestone
branches on "any section with count > 0", so the new category flows
through the existing prompt.

The resolved/unresolved predicate is NOT reimplemented. uat.cjs already
exports parseDeferredItems, which owns the parsing rule (entries under a
`## Deferred Items` level-2 heading, else the whole file fail-safe;
RESOLVED only on an explicit case-insensitive `status: resolved` field).
The scanner requires it lazily, inside the scan, so audit-command-router's
property that a route never loads the module it does not need is
preserved. Two readers of one file sharing one predicate is the point —
duplicating the inequality is how they drift into disagreeing about what
"open" means.

Regression test proves fail-first: 9 of its 10 cases go red against the
pre-change tree. The tenth is the deliberate no-regression boundary (a
clean tree emits no section) and is green both ways.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015TQU48ETJjEmGLJjA6hdQ4

* docs(#2646): document the pre-close artifact audit and its nine categories

The /gsd-complete-milestone entry did not mention the audit at all, so
the gate that can stop a close was undocumented — and this change adds a
category to it. Tabulates all nine with their source artifact and what
makes each one "open", plus the [R]/[A]/[C] outcomes.

Also disambiguates the one genuinely confusing thing: the per-phase
deferred-items.md scanned here is NOT the `## Deferred Items` section the
[A] path writes into STATE.md. Same name, different artifact, opposite
ends of the flow.

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015TQU48ETJjEmGLJjA6hdQ4

* chore(#2646): backfill changeset pr number

Co-Authored-By: Claude Opus 5 <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_015TQU48ETJjEmGLJjA6hdQ4

---------

Co-authored-by: Claude Opus 5 <noreply@anthropic.com>
Co-authored-by: Tom Boucher <trekkie@nomorestars.com>

.changeset/deferred-items-reach-milestone-close.md

@@ -0,0 +1,5 @@
+---
+type: Added
+pr: 2983
+---
+**Unresolved `deferred-items.md` entries now reach the milestone-close audit.** `auditOpenArtifacts` gains `deferred_items` as a ninth scanned category, so an out-of-scope discovery a phase agent correctly recorded rather than fixed surfaces in `/gsd-complete-milestone`'s pre-close report alongside the other eight, and the existing `[R] Resolve / [A] Acknowledge / [C] Cancel` prompt applies to it. #2287 made the file readable at the phase boundary (`audit-uat`, `/gsd-progress` check 7); one boundary up it was still invisible, and phase directories archive to `milestones/vX.Y-phases/` by default (#1871), so an unresolved entry left the live tree without ever being triaged. The resolved/unresolved predicate is not reimplemented — the scanner lazily requires `uat.cjs`'s exported `parseDeferredItems`, so both boundaries agree by construction about what "open" means. **Behaviour change worth noting before you upgrade:** a project carrying unresolved deferred items will now see the `[R]/[A]/[C]` prompt at milestone close where close previously proceeded silently. That is the intended correction, but it surfaces pre-existing debt on the first run. (#2646)