vigiles 14.9.0 → 14.10.1

package/dist/audit-verdict.js

@@ -229,13 +229,25 @@ function buildSentence(input, pointsToNextGrade, fixesToNextGrade) {
     if (score.empty) {
         return "No loadable harness surface — nothing to grade yet.";
     }
-    // Already an A: nothing is blocking the grade.
+    // Already an A: nothing is BLOCKING the grade — but "A" is not "flawless". Only
+    // claim "clean" at a real 100 (zero graded deductions); when graded findings
+    // pulled the score below 100, name them so the verdict never contradicts a
+    // finding the report shows (the superpowers "92 but structurally clean" bug).
     if (pointsToNextGrade === null) {
-        if (recommendations.length === 0) {
-            return "A — nothing blocking; the harness is structurally clean.";
+        if (score.overall === 100) {
+            return "A — clean. Nothing blocking, nothing flagged.";
         }
-        const n = recommendations.length;
-        return `A — nothing blocking the grade; ${numberWord(n)} deterministic ${fixNoun(n)} would harden it further.`;
+        if (recommendations.length > 0) {
+            const n = recommendations.length;
+            return `A — nothing blocks the grade; ${numberWord(n)} deterministic ${fixNoun(n)} would harden it further.`;
+        }
+        // Graded findings, but no auto-fix maps to them — name the dominant one so an
+        // A never reads as flawless while the report shows a real finding.
+        const dom = dominantDeduction(report);
+        if (dom !== null) {
+            return `A — nothing blocks the grade, but ${String(dom.n)} ${pluralizeLabel(dom.n, dom.label)} ${dom.n === 1 ? "is" : "are"} flagged — worth fixing.`;
+        }
+        return "A — nothing blocks the grade.";
     }
     // The next band's grade is gradeFor(its FLOOR); the floor is base + the gap.
     const nextGrade = (0, score_core_js_1.gradeFor)(score.overall + pointsToNextGrade);

package/package.json

@@ -1,6 +1,6 @@
 {
   "name": "vigiles",
-  "version": "14.9.0",
+  "version": "14.10.1",
   "description": "Lint & test the harness your AI agent runs on — verify the references in your CLAUDE.md / AGENTS.md and test that your hooks and skills actually work.",
   "keywords": [
     "claude-code",