fix(aidd-dev): a review never checks a runtime criterion from the code

plugins/aidd-dev/skills/05-review/actions/03-review-functional.md

@@ -4,7 +4,7 @@ Trace the diff against the plan's acceptance criteria.
 
 ## Input
 
-The plan, its phase files, and the diff prepare resolved.
+The plan, its phase files, any run recorded beside them, and the diff prepare resolved.
 
 ## Output
 
@@ -18,16 +18,17 @@ The round's criteria and score.
    - A phase whose `## Architecture projection` files the diff leaves untouched: out of the diff, no box. A phase projecting no file is traced, never out of the diff.
    - Met by the files as the diff leaves them, an added line or a context line: checked.
    - Unmet, partial, its subject absent, or a name echoing it with no behaviour behind: unchecked, never a finding.
-   - Its subject there, its property one no diff could ever show: checked, not-applicable.
-3. **Score.** Count met + unmet + out of the diff = the plan's whole criteria total.
-   - A not-applicable criterion counts as met.
+   - Its subject there, its outcome one only a run could show, what a screen displays, a timing, an outside system's answer: unverified, never checked from the code.
+   - A run recorded beside the plan on the reviewed head shows it passing: checked, citing the record; failing: unchecked; recorded on another head: unverified.
+3. **Score.** Count met + unmet + unverified + out of the diff = the plan's whole criteria total.
 
 ## Test
 
 | Case | Pass |
 | --- | --- |
-| A plan of 6 criteria, the diff touching one phase of 2 with 1 met | `1/6 met, 1 unmet, 4 out of the diff`, the untouched phases out of the diff in plan order |
+| A plan of 6 criteria, the diff touching one phase of 2 with 1 met | `1/6 met, 1 unmet, 0 unverified at runtime, 4 out of the diff`, the untouched phases out of the diff in plan order |
 | A criterion unmet, its subject absent, or a name echoing it with no behaviour behind | unchecked, naming the gap, never a finding |
-| A criterion whose property no diff could show | checked, not-applicable, counted as met |
+| A criterion only a run could show, and no run recorded | unverified, never checked |
+| The same, a recorded run passing it | checked, citing the record |
 | No plan and no criteria given | not scored, and no criteria |
 | A second round over the same plan | every box follows the current diff, none carried from the earlier round |

plugins/aidd-dev/skills/05-review/assets/review-template.md

@@ -10,13 +10,13 @@
 - Diff: `{{base}}...{{head}}`{{ plus the working tree, when it held changes}}
 - Axes: {{the axes that ran, from code, functional, relevancy}}
 - Verdict: {{verdict}}
-- Score: {{n_met}}/{{n_plan}} met, {{n_unmet}} unmet, {{n_out}} out of the diff, or not scored
+- Score: {{n_met}}/{{n_plan}} met, {{n_unmet}} unmet, {{n_unverified}} unverified at runtime, {{n_out}} out of the diff, or not scored
 
 ### Criteria
 
-- [x] {{criterion met}} — {{file:line}}
+- [x] {{criterion met}} — {{file:line, or the run record}}
 - [ ] {{criterion unmet}} — {{gap}}
-- [x] {{criterion no diff could show}} — not-applicable, {{why}}
+- Unverified at runtime: {{criterion}} — {{what a run would show}}
 - Out of the diff: {{phase-name}} ({{n}} criteria)
 
 ### Findings

plugins/aidd-dev/skills/05-review/references/report-contract.md

@@ -23,5 +23,5 @@ Where a review lands, what it reviews, and how a round joins it.
 
 | Rule | Why |
 | --- | --- |
-| Append a round, derived from the plan and the current diff alone | a round states what this diff shows, and nothing it inherited |
+| Append a round, derived from the plan, the current diff and a run recorded on its head alone | a round states what this candidate shows, and nothing it inherited |
 | Count the `## Round` headings, never read their bodies | an earlier round's boxes describe a diff that no longer exists, and copying one is the defect |

plugins/aidd-dev/skills/05-review/references/review-rubric.md

@@ -21,4 +21,5 @@ One overall verdict, the strictest across the axes run.
 | Case | Counts as |
 | --- | --- |
 | A criterion out of the diff | neither unchecked nor blocking |
+| A criterion unverified at runtime | in scope, neither unchecked nor blocking |
 | A round without functional | a verdict from its findings alone |

scripts/__tests__/a-review-appends-a-round-and-scores-the-whole-plan.test.js

@@ -49,7 +49,7 @@ function inSkill(file) {
 describe("a review appends a round and scores the whole plan", () => {
   it("a round is appended, and an earlier one is never read back", () => {
     const reference = read("references/report-contract.md");
-    assert.match(reference, /Append a round, derived from the plan and the current diff alone/u);
+    assert.match(reference, /Append a round, derived from the plan, the current diff and a run recorded on its head alone/u);
     assert.match(
       reference,
       /Count the `## Round` headings, never read their bodies/u,
@@ -205,10 +205,21 @@ describe("a review appends a round and scores the whole plan", () => {
     );
   });
 
-  it("a criterion no diff could show is not-applicable, not a gap", () => {
+  it("a criterion only a run could show is unverified, never met on the code alone", () => {
+    const trace = step("actions/03-review-functional.md", "Trace");
+    assert.match(trace, /its outcome one only a run could show[^\n]*: unverified, never checked from the code/u);
+    assert.match(trace, /only a run could show, what a screen displays, a timing, an outside system's answer:/u, "a display is a run's to show, even beside a timing");
+    assert.match(trace, /A run recorded beside the plan on the reviewed head shows it passing: checked, citing the record; failing: unchecked; recorded on another head: unverified/u);
+    const strays = everyMarkdown(SKILL, [])
+      .filter((file) => /not-applicable/u.test(fs.readFileSync(file, "utf8")))
+      .map(inSkill);
+    assert.deepEqual(strays, [], "one state for a criterion the code cannot prove");
+  });
+
+  it("an unverified criterion neither blocks nor asks for changes", () => {
     assert.match(
-      step("actions/03-review-functional.md", "Trace"),
-      /Its subject there, its property one no diff could ever show: checked, not-applicable/u,
+      read("references/review-rubric.md"),
+      /^\| A criterion unverified at runtime \| in scope, neither unchecked nor blocking \|$/mu,
     );
   });
 
@@ -265,14 +276,13 @@ describe("a review appends a round and scores the whole plan", () => {
   it("the template renders every form a round can hold", () => {
     const template = read("assets/review-template.md");
     assert.match(template, /or not scored/u);
-    assert.match(template, /not-applicable, \{\{why\}\}/u);
+    assert.match(template, /^- Unverified at runtime: \{\{criterion\}\} — \{\{what a run would show\}\}$/mu);
   });
 
   it("the score counts over the whole plan, and the template holds its form", () => {
     const score = step("actions/03-review-functional.md", "Score");
     assert.match(score, /whole criteria total/u, "the count is over the plan, not over what is in scope");
-    assert.match(score, /met \+ unmet \+ out of the diff = the plan's whole criteria total/u, "and the three numbers are checkable against it");
-    assert.match(score, /counts as met/u, "and a not-applicable criterion is counted");
+    assert.match(score, /met \+ unmet \+ unverified \+ out of the diff = the plan's whole criteria total/u, "and the four numbers are checkable against it");
     assert.doesNotMatch(score, /\{\{/u, "a rule names no template placeholder");
     assert.match(
       step("actions/03-review-functional.md", "Read"),
@@ -281,7 +291,7 @@ describe("a review appends a round and scores the whole plan", () => {
     );
     assert.match(
       read("assets/review-template.md"),
-      /\{\{n_met\}\}\/\{\{n_plan\}\} met, \{\{n_unmet\}\} unmet, \{\{n_out\}\} out of the diff/u,
+      /\{\{n_met\}\}\/\{\{n_plan\}\} met, \{\{n_unmet\}\} unmet, \{\{n_unverified\}\} unverified at runtime, \{\{n_out\}\} out of the diff/u,
       "the form is rendered by the template, not spelled out in a rule",
     );
   });