feat(review): pattern column on the close-out acceptance-criteria rollup

plugins/review/.claude-plugin/plugin.json

@@ -1,7 +1,7 @@
 {
   "$schema": "https://json.schemastore.org/claude-code-plugin-manifest.json",
   "name": "review",
-  "version": "0.26.19",
+  "version": "0.26.20",
   "description": "Code-review toolkit: six read-only reviewer agents (code, security, architecture, doc drift, build/test/lint, CI-log audit) plus orchestration skills \u2014 quality gate, fan-out, and CI lane commands (/review:code-review, /review:security-review) for org reusable workflows.",
   "author": {
     "name": "Melodic Software",

plugins/review/CHANGELOG.md

@@ -3,6 +3,23 @@
 All notable changes to the `review` plugin are documented here. Format follows
 [Keep a Changelog](https://keepachangelog.com/en/1.1.0/); this plugin uses semantic versioning.
 
+## [0.26.20]
+
+### Added
+
+- **`quality-gate`:** close-out mode's acceptance-criteria rollup gains a requirement-pattern
+  column when any of the container's retrieved criteria opens with a bracketed EARS tag. The cell
+  carries one of exactly five names, `ubiquitous`, `event-driven`, `state-driven`,
+  `unwanted-behaviour`, `optional-feature`, matching what the planning surfaces emit; a bracket
+  holding anything else leaves the cell empty rather than echoing raw text. Detection is a leading
+  bracket holding one of those names, so a checklist marker (`- [ ]`) is not read as a tag and no
+  flag, lever, or convention key is read at all. Every criterion still gets a
+  row, so a partially tagged set shows its untagged rows with an empty pattern cell, and a set
+  with no tag renders exactly as before with no extra column. Step 2's extraction bullet now says
+  to keep the criterion line whole so the tag survives the read. The verdict vocabulary
+  (`delivered` / `partial` / `missing` / `unverifiable`) and the blocking rule (`missing` or
+  `wrong` keeps the container open) are unchanged.
+
 ## [0.26.19]
 
 ### Changed

plugins/review/README.md

@@ -30,7 +30,9 @@ Invoke via `@review:<agent>` or let Claude delegate.
   `security`, `spec` (spec-fidelity: did the change deliver what the originating item, plan, or
   brief asked for), `close-out` (the same fidelity lens at spec-container scale, one cumulative
   pass over everything a container shipped, across however many PRs, against the container's own
-  body; derives its own diff basis per execution shape), `downstream` (what the change breaks
+  body; derives its own diff basis per execution shape, and its acceptance-criteria rollup gains a
+  requirement-pattern column when the container's criteria carry bracketed EARS tags),
+  `downstream` (what the change breaks
   outside its own diff: callers, serialization boundaries, cross-service consumers), `pr`,
   `criteria`, `slice <name>`, `restatement`.
 - **`/review:fanout [mode]`**. Breadth review: fans out across the

plugins/review/skills/quality-gate/context/close-out.md

@@ -93,7 +93,9 @@ inside the verbatim fence [spec.md](spec.md) "Step 2" specifies.
 
 From the body, extract the three things the rest of this mode needs:
 
-- the **acceptance criteria** — the checklist the cumulative verdict is rendered against
+- the **acceptance criteria** — the checklist the cumulative verdict is rendered against. Keep each
+  criterion line whole rather than stripping it to prose: a leading bracketed tag is part of the
+  line, and Step 6 reads it
 - any **scope statement** (an `## Out of scope` section, an acceptance-criteria list read as
   exhaustive) — without one, `scope-creep` is not reachable at all, per [spec.md](spec.md)
   "Finding classes"
@@ -369,7 +371,20 @@ plus, above it:
   `sub-item → PR → oid`, so the basis is auditable rather than asserted; name the rung, and say
   outright when it was the heuristic scan
 - the **acceptance-criteria rollup** — every criterion with its `delivered` / `partial` /
-  `missing` / `unverifiable` verdict
+  `missing` / `unverifiable` verdict, and, **when any criterion retrieved in Step 2 opens with a
+  bracketed EARS tag, one more column naming that criterion's requirement pattern**. That column is
+  what makes the shape of a tagged set legible: that every `unwanted-behaviour` criterion came back
+  `unverifiable` while the `event-driven` ones were all `delivered` is a fact the verdict column
+  alone cannot show. The pattern cell carries one of exactly five names, `ubiquitous`,
+  `event-driven`, `state-driven`, `unwanted-behaviour`, `optional-feature`, or nothing at all; a
+  bracket holding anything else is an untagged criterion that looks tagged, so the cell stays empty
+  rather than echoing the raw text. Detection is a leading bracket holding one of those five names
+  in the criteria as retrieved, nothing else: a checklist marker (`- [ ]`) is a bracket and not a
+  tag, and no flag, no lever, and no convention key is read here. **Every criterion
+  still gets a row.** A partially tagged set leaves the untagged rows' pattern cell empty rather
+  than dropping them, which is the invisibility this rollup exists to remove; a set carrying no tag
+  at all renders exactly as it does today, with no pattern column. The column names the pattern and
+  changes no verdict
 - the **`no-code` sub-items** — those the provider confirms closed without a PR, each with the
   closing comment its criteria were judged against. These are journey coverage, not gaps.
 - any sub-item whose shipping commit could not be resolved (`unresolved`), listed as a coverage gap