chore(release): prepare v3.13.5

CHANGELOG.md

@@ -5,6 +5,54 @@ All notable changes to the Agentic QE project will be documented in this file.
 The format is based on [Keep a Changelog](https://keepachangelog.com/en/1.0.0/),
 and this project adheres to [Semantic Versioning](https://semver.org/spec/v2.0.0.html).
 
+## [3.13.5] - 2026-08-03
+
+Quality decisions and generated tests now carry evidence that reflects what AQE
+actually measured or executed, while initialization and embedding workflows are
+more reliable across supported environments.
+
+### Added
+
+- **Measured quality-gate evidence** ([#596]). CLI and MCP quality evaluation
+  now share canonical thresholds and require fresh, attributed measurements for
+  coverage, passing tests, critical bugs, code smells, security findings,
+  technical debt, and duplication. Missing, malformed, or stale evidence fails
+  closed instead of silently receiving a fabricated passing value.
+- **Optional status-line installation** ([#591]). `aqe init --no-statusline`
+  skips AQE's status line, removes an existing AQE-owned status line during an
+  upgrade, and preserves a project-authored status line.
+
+### Fixed
+
+- **Generated tests are trustworthy and executable** ([#577]). Test generation
+  preserves framework imports, validates evidence before reporting success, and
+  no longer treats structurally incomplete output as a passing quality result.
+- **Node's built-in test runner is honored end to end** ([#592]). Test discovery
+  filters Node test files correctly, execution uses the selected framework, and
+  retry confidence remains bounded.
+- **Batch embedding semantics are consistent** ([#585]). Mixed batches send only
+  semantic content to the embedding provider, nonsemantic entries receive zero
+  vectors in their original positions, and all-nonsemantic batches avoid
+  provider initialization entirely.
+- **HNSW initialization works in native ESM** ([#586]). Runtime dependency
+  loading no longer assumes a CommonJS global `require`.
+- **Learning-engine tests use isolated persistence** ([#588]), preventing shared
+  worker databases from accumulating duplicate genesis data and causing long
+  test runs.
+
+### Changed
+
+- Refreshed bundled Claude helpers and development-only Ruflo dependencies,
+  including upstream security and compatibility fixes.
+
+[#577]: https://github.com/proffesor-for-testing/agentic-qe/issues/577
+[#585]: https://github.com/proffesor-for-testing/agentic-qe/issues/585
+[#586]: https://github.com/proffesor-for-testing/agentic-qe/issues/586
+[#588]: https://github.com/proffesor-for-testing/agentic-qe/issues/588
+[#591]: https://github.com/proffesor-for-testing/agentic-qe/issues/591
+[#592]: https://github.com/proffesor-for-testing/agentic-qe/issues/592
+[#596]: https://github.com/proffesor-for-testing/agentic-qe/issues/596
+
 ## [3.13.4] - 2026-08-02
 
 Codex users now receive a complete, upgradeable Agentic QE workflow instead of

assets/skills/skills-manifest.json

@@ -940,7 +940,7 @@
   },
   "metadata": {
     "generatedBy": "Agentic QE Fleet",
-    "fleetVersion": "3.13.4",
+    "fleetVersion": "3.13.5",
     "manifestVersion": "1.4.0",
     "lastUpdated": "2026-04-13T00:00:00.000Z",
     "contributors": [

docs/releases/README.md

@@ -4,6 +4,7 @@ All Agentic QE release notes organized by version.
 
 | Version | Date | Highlights |
 |---------|------|------------|
+| [v3.13.5](v3.13.5.md) | 2026-08-03 | Measured quality gates and trustworthy cross-framework test execution. |
 | [v3.13.4](v3.13.4.md) | 2026-08-02 | Reliable Codex fleet setup, upgrades, hooks, and verification. |
 | [v3.13.3](v3.13.3.md) | 2026-07-29 | Learning capture no longer stops silently under `AQE_DISABLE_WAL` (it now refuses to write in an unsafe journal mode and names what is holding the DB), and QE-Court's shipped panel no longer violates its own `writerIsNeverJuror` rule (#576). |
 | [v3.13.2](v3.13.2.md) | 2026-07-24 | Stops hook-driven `brain.rvf`/`patterns.rvf` corruption loops by closing native stores after every hook and protecting same-process live locks (#574). |

docs/releases/v3.13.5.md

@@ -0,0 +1,61 @@
+# v3.13.5 Release Notes
+
+**Release Date:** 2026-08-03
+
+## Highlights
+
+AQE now bases quality decisions on fresh, attributed measurements and produces
+more trustworthy, executable tests across Jest, Vitest, and Node's built-in test
+runner. Initialization and embeddings are also more reliable in real projects.
+
+## Added
+
+- Canonical quality evidence for coverage, passing tests, bugs, code smells,
+  security findings, technical debt, and duplication.
+- Shared CLI and MCP quality thresholds that fail closed when measurements are
+  missing, malformed, or stale.
+- `aqe init --no-statusline` for installations that do not want AQE's status
+  line, without disturbing a project-authored status line.
+
+## Fixed
+
+- Generated tests retain the correct framework imports and must satisfy
+  evidence-backed validation before AQE reports success.
+- Node test discovery, execution, and retry behavior now honor the selected
+  framework consistently.
+- Mixed embedding batches preserve output order while excluding nonsemantic
+  content from provider calls; entirely nonsemantic batches make no provider
+  call.
+- HNSW runtime loading works when AQE is imported as native ESM.
+- Learning-engine unit tests no longer share persistence across test cases.
+
+## Changed
+
+- Refreshed bundled Claude helpers.
+- Updated development-only Ruflo UI and tooling dependencies, including upstream
+  security fixes.
+
+## Upgrade notes
+
+No configuration migration is required. Existing installations can upgrade and
+refresh their managed assets normally:
+
+```bash
+npx agentic-qe init --auto --upgrade
+```
+
+Projects that want to remove only AQE's managed status line can run:
+
+```bash
+npx agentic-qe init --auto --upgrade --no-statusline
+```
+
+Project-authored status-line configuration is preserved.
+
+Tracking: [#577](https://github.com/proffesor-for-testing/agentic-qe/issues/577),
+[#585](https://github.com/proffesor-for-testing/agentic-qe/issues/585),
+[#586](https://github.com/proffesor-for-testing/agentic-qe/issues/586),
+[#588](https://github.com/proffesor-for-testing/agentic-qe/issues/588),
+[#591](https://github.com/proffesor-for-testing/agentic-qe/issues/591),
+[#592](https://github.com/proffesor-for-testing/agentic-qe/issues/592), and
+[#596](https://github.com/proffesor-for-testing/agentic-qe/issues/596).

package-lock.json

@@ -1,12 +1,12 @@
 {
   "name": "agentic-qe",
-  "version": "3.13.4",
+  "version": "3.13.5",
   "lockfileVersion": 3,
   "requires": true,
   "packages": {
     "": {
       "name": "agentic-qe",
-      "version": "3.13.4",
+      "version": "3.13.5",
       "hasInstallScript": true,
       "license": "MIT",
       "dependencies": {

package.json

@@ -1,6 +1,6 @@
 {
   "name": "agentic-qe",
-  "version": "3.13.4",
+  "version": "3.13.5",
   "description": "Agentic Quality Engineering V3 - Domain-Driven Design Architecture with 13 Bounded Contexts, O(log n) coverage analysis, ReasoningBank learning, 60 specialized QE agents, mathematical Coherence verification, deep Claude Flow integration",
   "type": "module",
   "main": "./dist/index.js",
@@ -71,7 +71,7 @@
     "test:unit:fast": "vitest run --exclude='**/qe-reasoning-bank.test.ts' --exclude='**/qe-reasoning-bank-feedback-loop.test.ts' --exclude='**/aqe-learning-engine.test.ts' --exclude='**/aqe-learning-engine-patterns.test.ts' tests/unit/adapters tests/unit/shared tests/unit/cli tests/unit/learning tests/unit/kernel tests/unit/workers tests/unit/routing tests/unit/strange-loop tests/unit/sync tests/unit/feedback tests/unit/error-paths tests/unit/early-exit tests/unit/causal-discovery tests/unit/neural-optimizer tests/unit/test-scheduling tests/unit/logging tests/unit/validation tests/unit/memory tests/unit/performance tests/unit/scripts tests/unit/planning",
     "test:unit:heavy": "vitest run tests/unit/coordination tests/unit/domains tests/unit/integrations tests/unit/optimization tests/unit/init --exclude='**/browser/**' --exclude='**/*.e2e.test.ts' --exclude='**/vibium/**' --exclude='**/browser-swarm-coordinator.test.ts'",
     "test:unit:mcp": "vitest run tests/unit/mcp --exclude='**/mcp/handlers/domain-handlers.test.ts' --fileParallelism=false",
-    "test:ci": "vitest run --exclude='**/browser/**' --exclude='**/*.e2e.test.ts' --exclude='**/vibium/**' --exclude='**/integration/browser/**' --exclude='**/browser-swarm-coordinator.test.ts' --exclude='**/mcp/handlers/domain-handlers.test.ts'",
+    "test:ci": "vitest run --exclude='**/browser/**' --exclude='**/*.e2e.test.ts' --exclude='**/vibium/**' --exclude='**/integration/browser/**' --exclude='**/browser-swarm-coordinator.test.ts' --exclude='**/mcp/handlers/domain-handlers.test.ts' --exclude='**/fixtures/init-corpus/**'",
     "test:e2e": "vitest run tests/integration/browser --testTimeout=120000",
     "test:safe": "NODE_OPTIONS='--max-old-space-size=768 --expose-gc' vitest run --maxForks=1",
     "test:dev": "npm run test:unit:fast",

tests/integration/ruvector/native-hnsw-real-fixture.test.ts

@@ -13,7 +13,7 @@
  * in production, written by sentence-transformers/all-MiniLM-L6-v2-style
  * models) and verifies:
  *
- *   1. Self-query returns id=self with score 1.0 (exact-match recall)
+ *   1. Self-query returns an exact vector match with score 1.0
  *   2. Top-10 recall vs brute-force ground truth = 100%
  *   3. Repeated for several random query vectors to rule out one-off luck
  *   4. Vectors persisted in memory.db deserialize correctly into the backend
@@ -183,7 +183,7 @@ describe('NativeHnswBackend — real qe-kernel fixture (#399)', () => {
   );
 
   it.runIf(fixture !== null && fixture.length >= 100)(
-    'should hit recall@10 >= 0.9 and top-1 == self on real qe-kernel embeddings',
+    'should hit recall@10 >= 0.9 with an exact top match on real qe-kernel embeddings',
     () => {
       const pool = fixture!;
       const backend = new NativeHnswBackend({
@@ -209,13 +209,12 @@ describe('NativeHnswBackend — real qe-kernel fixture (#399)', () => {
       //     them based on entry-point luck. 90-100% recall@10 is the
       //     standard approximate-HNSW guarantee.
       //
-      // The TOP-1 self-match assertion is the strict bug-detector — it's
-      // mathematically impossible to fail unless HNSW is fundamentally
-      // broken (which is exactly what @ruvector/router 0.1.28 was: it
-      // returned random non-neighbors with recall@10 = 0% to 10%). Anything
-      // above 0.5 recall@10 here would already be a clear pass. We hold
-      // the bar at 0.9 to catch tuning regressions while accepting the
-      // legitimate plateau-region behavior of real embeddings.
+      // The TOP-1 exact-vector assertion is the strict bug-detector. Several
+      // real fixture rows can contain byte-identical embeddings, so requiring
+      // the queried row's numeric id would make the result depend on arbitrary
+      // tie-breaking. @ruvector/router 0.1.28 returned random non-neighbors
+      // with recall@10 = 0% to 10%; requiring an exact top result still catches
+      // that failure while accepting indistinguishable duplicate vectors.
       const RECALL_FLOOR = 0.9;
 
       for (const queryIdx of [
@@ -234,7 +233,8 @@ describe('NativeHnswBackend — real qe-kernel fixture (#399)', () => {
         const recallAt10 =
           [...groundTruthIds].filter((id) => hitIds.has(id)).length / groundTruthIds.size;
 
-        expect(results[0].id).toBe(queryIdx); // self MUST be top-1 (exact-match guarantee)
+        expect(pool[results[0].id].vector).toEqual(queryVector);
+        expect(results[0].score).toBeGreaterThanOrEqual(0.999);
         expect(recallAt10).toBeGreaterThanOrEqual(RECALL_FLOOR);
       }