agentic-qe 3.9.12 → 3.9.14

package/.claude/agents/v3/qe-coverage-specialist.md

@@ -11,7 +11,7 @@ domain: coverage-analysis
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-fleet-commander.md

@@ -8,7 +8,7 @@ domain: cross-domain
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-pentest-validator.md

@@ -1,14 +1,16 @@
 ---
 name: qe-pentest-validator
 version: "3.6.0"
-updated: "2026-04-12"
+updated: "2026-04-17"
 description: Graduated exploit validation with parallel vulnerability pipelines, browser-based attack execution, and "No Exploit, No Report" quality gate
 v2_compat: null
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 advisor:
   enabled: true
   provider: claude
-  model: claude-opus-4-6
+  model: claude-opus-4-7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-queen-coordinator.md

@@ -14,7 +14,7 @@ dependencies:
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-risk-assessor.md

@@ -8,7 +8,7 @@ v3_new: true
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-root-cause-analyzer.md

@@ -19,7 +19,7 @@ dependencies:
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-security-auditor.md

@@ -1,14 +1,16 @@
 ---
 name: qe-security-auditor
 version: "3.0.0"
-updated: "2026-04-12"
+updated: "2026-04-17"
 description: Security audit specialist with OWASP coverage, compliance validation, and remediation workflows
 v2_compat: null # New in v3
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 advisor:
   enabled: true
   provider: claude
-  model: claude-opus-4-6
+  model: claude-opus-4-7
   max_uses: 3
   redact: strict
 ---

package/.claude/agents/v3/qe-security-scanner.md

@@ -1,10 +1,12 @@
 ---
 name: qe-security-scanner
 version: "3.0.0"
-updated: "2026-01-10"
+updated: "2026-04-17"
 description: Comprehensive security scanning with SAST, DAST, dependency scanning, and secrets detection
 v2_compat: qe-security-scanner
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 dependencies:
   agents:
     - name: qe-dependency-mapper

package/.claude/agents/v3/qe-test-architect.md

@@ -11,7 +11,7 @@ domain: test-generation
 advisor:
   enabled: true                    # ADR-092 Phase 0 target agent
   provider: openrouter             # Phase 0 default; any HybridRouter provider supported
-  model: anthropic/claude-opus-4   # Strong reasoning model via OpenRouter
+  model: anthropic/claude-opus-4.7   # Strong reasoning model via OpenRouter
   max_uses: 3                      # Per-task advisor call cap
   budget_usd_per_task: 0.05        # Per-call budget ceiling
   required: false                  # Phase 5 flips this to true for quality-gate enforcement

package/.claude/commands/aqe-costs.md

@@ -145,7 +145,7 @@ By Provider:
     Tokens: 891,234
     Cost: $4.5678
     Avg Cost/Request: $0.018644
-    Top Model: claude-sonnet-4-5-20250929
+    Top Model: claude-sonnet-4-6
 
   ☁️ openrouter:
     Requests: 111
@@ -206,10 +206,10 @@ By Provider:
       "totalTokens": 891234,
       "totalCost": 4.5678,
       "avgCostPerRequest": 0.018644,
-      "topModel": "claude-sonnet-4-5-20250929",
+      "topModel": "claude-sonnet-4-6",
       "modelCounts": {
-        "claude-sonnet-4-5-20250929": 187,
-        "claude-3-5-haiku-20241022": 58
+        "claude-sonnet-4-6": 187,
+        "claude-haiku-4-5-20251001": 58
       }
     },
     "openrouter": {
@@ -245,7 +245,7 @@ Recent Requests (Last 20):
   Tokens: 1,234 input / 567 output = 1,801 total
   Cost: $0.0000
 
-[2025-12-15T23:57:23Z] anthropic/claude-sonnet-4-5-20250929
+[2025-12-15T23:57:23Z] anthropic/claude-sonnet-4-6
   Agent: qe-quality-gate
   Task: quality-check-456
   Tokens: 3,456 input / 1,789 output = 5,245 total
@@ -321,7 +321,7 @@ tracker.trackRequest({
 // Track cloud inference
 tracker.trackRequest({
   provider: 'anthropic',
-  model: 'claude-sonnet-4-5-20250929',
+  model: 'claude-sonnet-4-6',
   tokens: {
     inputTokens: 2000,
     outputTokens: 1000,

package/.claude/skills/.validation/schemas/skill-eval.schema.json

@@ -26,6 +26,9 @@
       "items": {
         "type": "string",
         "enum": [
+          "claude-opus-4-7",
+          "claude-sonnet-4-6",
+          "claude-haiku-4-5",
           "claude-opus-4-5",
           "claude-sonnet-4",
           "claude-3.5-sonnet",

package/.claude/skills/.validation/test-data/sample-output.json

@@ -60,7 +60,7 @@
     "executionTimeMs": 1250,
     "toolsUsed": ["eslint", "typescript", "custom-analyzer"],
     "agentId": "qe-test-architect",
-    "modelUsed": "claude-sonnet-4",
+    "modelUsed": "claude-sonnet-4-6",
     "inputHash": "sha256:abc123def456",
     "parentTaskId": "task-789"
   },

package/.claude/skills/a11y-ally/SKILL.md

@@ -1230,7 +1230,7 @@ const client = new Anthropic();
 
 const imageData = fs.readFileSync('/tmp/a11y-work/frames/frame_01.jpg').toString('base64');
 const response = await client.messages.create({
-  model: 'claude-sonnet-4-20250514',
+  model: 'claude-sonnet-4-6',
   max_tokens: 500,
   messages: [{
     role: 'user',

package/.claude/skills/accessibility-testing/evals/accessibility-testing.yaml

@@ -28,7 +28,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (ensure minimum quality)
+  - claude-haiku-4-5          # Fast model (ensure minimum quality)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/api-testing-patterns/evals/api-testing-patterns.yaml

@@ -30,8 +30,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4       # Primary model (high accuracy expected)
-  - claude-3-haiku        # Fast model (minimum quality bar)
+  - claude-sonnet-4-6        # Primary model (high accuracy expected)
+  - claude-haiku-4-5          # Fast model (minimum quality bar)
   - gpt-4o                # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/chaos-engineering-resilience/evals/chaos-engineering-resilience.yaml

@@ -35,7 +35,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/contract-testing/evals/contract-testing.yaml

@@ -30,8 +30,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4       # Primary model (high accuracy expected)
-  - claude-3-haiku        # Fast model (minimum quality bar)
+  - claude-sonnet-4-6        # Primary model (high accuracy expected)
+  - claude-haiku-4-5          # Fast model (minimum quality bar)
   - gpt-4o                # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/database-testing/evals/database-testing.yaml

@@ -33,8 +33,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4     # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-sonnet-4-6      # Primary model (high accuracy expected)
+  - claude-haiku-4-5         # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/localization-testing/evals/localization-testing.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
 
 # =============================================================================
 # MCP Integration Configuration

package/.claude/skills/n8n-trigger-testing-strategies/evals/n8n-trigger-testing-strategies.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet
-  - claude-3-haiku
+  - claude-haiku-4-5
 
 # =============================================================================
 # MCP Integration Configuration

package/.claude/skills/performance-testing/evals/performance-testing.yaml

@@ -35,7 +35,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/.claude/skills/quality-metrics/evals/quality-metrics.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet
-  - claude-3-haiku
+  - claude-haiku-4-5
 
 # =============================================================================
 # MCP Integration Configuration

package/.claude/skills/skills-manifest.json

@@ -939,7 +939,7 @@
   },
   "metadata": {
     "generatedBy": "Agentic QE Fleet",
-    "fleetVersion": "3.9.12",
+    "fleetVersion": "3.9.14",
     "manifestVersion": "1.4.0",
     "lastUpdated": "2026-04-13T00:00:00.000Z",
     "contributors": [

package/.claude/skills/testability-scoring/evals/testability-scoring.yaml

@@ -29,8 +29,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4-20250514    # Primary model (high accuracy expected)
-  - claude-3-haiku              # Fast model (minimum quality threshold)
+  - claude-sonnet-4-6    # Primary model (high accuracy expected)
+  - claude-haiku-4-5                # Fast model (minimum quality threshold)
   - gpt-4o                      # Cross-vendor validation
 
 # =============================================================================

package/CHANGELOG.md

@@ -5,6 +5,51 @@ 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).
 
+## [3.9.14] - 2026-04-20
+
+**Security + supply-chain hardening.** Closes five P0 release blockers from the v3.9.13 QE audit: 15 critical runtime npm vulnerabilities, 79% tarball bloat, hardcoded retiring model IDs at tier 1, a broken lint harness, and a loose MCP contract. Tarball shipped size drops from 19.9 MB to 9.6 MB (-52%). Also tightens six MCP tool contracts, patches a command-injection path in `aqe learning repair`, and stops the telemetry workflow from push-forcing to protected `main`.
+
+### Fixed
+
+- **15 critical runtime vulnerabilities eliminated** — The `@xenova/transformers → @claude-flow/{browser,guidance,embeddings} → onnxruntime-web → onnx-proto → protobufjs` chain pulled in protobufjs `<7.5.5` (GHSA-xq3m-2v4x-88gg, CWE-94 arbitrary code execution). Pinned to `^7.5.5` via both `overrides` and `resolutions`. `npm audit --omit=dev` now reports zero critical/high issues.
+- **Command injection in `aqe learning repair`** — The `--file` path was shell-interpolated into three `execSync` calls. A crafted path could break out of shell quoting and execute arbitrary commands. Rewrote to `execFileSync` with explicit file-descriptor redirection and stdin streaming — no shell parsing of user input at all.
+- **`advisor_consult` MCP contract** — The ADR-092 advisor tool accepted empty/whitespace `agent` and `task` values and silently shelled out to `aqe llm advise` with placeholders. Now rejects missing or blank inputs before any child process spawn.
+- **Tier-1 model tier was hardcoded to a retiring model ID** — Six call sites (central constants plus five domain services: test-execution, contract-testing, chaos-resilience, learning-optimization, visual-accessibility) resolved tier 1 to `claude-3-haiku-20240307`, which will 404 after Haiku 3 retirement. Routed through `claude-haiku-4-5` (ADR-093).
+- **`npm run lint` was broken** — The script invoked eslint on both `src` and `tests`, but `.eslintrc.cjs` explicitly ignores `tests/`. ESLint aborted instead of linting anything. Narrowed the script to `eslint src --ext .ts` so the harness runs.
+- **`qcsd-production-trigger.yml` no longer pushes to protected `main`** — The post-publish telemetry job ran `git push` directly to `main`, which failed 8/10 times because of branch protection. Now pushes to a bot branch and opens a PR for maintainer review. Artifact upload continues to preserve the raw payload for 90 days regardless of PR merge.
+
+### Changed
+
+- **Tarball is 52% smaller** — `dist/cli/chunks/` accumulated stale code-split chunks across rebuilds (799 shipped, only 240 fresh). Build script now cleans the chunks directory before each build. Combined with lazy-loading `@faker-js/faker` (see below), shipped size drops from **19.9 MB → 9.6 MB** and unpacked from **88.5 MB → 48.2 MB**.
+- **`@faker-js/faker` is no longer bundled** — `test-data-generator.ts` previously static-imported faker at the module top, pulling the devDep into every shipped artefact. Now lazily loads via dynamic `import()` on first use. Faker is an optional runtime dep — users who invoke test-data generation must install it themselves (with a clear error message pointing them to `npm install --save-dev @faker-js/faker`).
+- **`@claude-flow/guidance` demoted to `optionalDependencies`** — Usage is already guarded by `try/await import/catch` fallback paths, and no newer stable version exists on npm. Keeping it installable for users who want it without forcing it on users who don't. See `docs/qe-reports-3-9-13/research-claude-flow-guidance-strategy.md`.
+- **Six MCP tools now mark their load-bearing params as required** — `agent_metrics(agentId)`, `coverage_analyze_sublinear(target)`, `accessibility_test(url)`, `defect_predict(target)`, `code_index(target)`, and `advisor_consult(agent, task)`. Clients that omit these parameters get a clear rejection instead of the tool running with empty placeholders.
+
+### Added
+
+- **v3.9.13 full-quality audit** — Twelve reports from an 11-agent QE swarm analysing code complexity, security, performance, test quality, SFDIPOT product factors, dependency health, API contracts, DDD architecture, accessibility, brutal honesty, and CI pipelines. Published at `docs/qe-reports-3-9-13/` alongside two follow-up research memos on the guidance strategy and agent-count recount.
+
+## [3.9.13] - 2026-04-17
+
+**Migrate the fleet to Claude Opus 4.7 / Sonnet 4.6 / Haiku 4.5 ahead of the June 15 Sonnet 4 retirement.** `xhigh` becomes the fleet-wide default effort level for better agentic-coding quality, and security agents stay pinned to Sonnet 4.6 until the Cyber Verification Program application clears. ([ADR-093](docs/implementation/adrs/ADR-093-opus-4-7-migration.md))
+
+### Added
+
+- **Opus 4.7, Sonnet 4.6, and Haiku 4.5 in the model registry** — New Anthropic models are first-class, with capability flags for adaptive thinking, `xhigh` effort, tokenizer version, and 1M context. Opus 4.7 is the default ADR-092 advisor model.
+- **`xhigh` effort level, fleet-wide** — Every qe-* agent now runs at `xhigh` by default (Anthropic's recommended starting point for agentic coding). Override per-agent via frontmatter, per-run via `QE_EFFORT_LEVEL`, or globally via `config/fleet-defaults.yaml`. Priority chain: env > YAML > frontmatter.
+- **Cyber Verification Program application draft** — Prepared at `docs/security/cyber-verification-application.md` for the four cyber-sensitive agents (`qe-pentest-validator`, `qe-security-scanner`, `qe-security-auditor`, `qe-security-reviewer`), which remain pinned to Sonnet 4.6 until enrollment clears.
+
+### Changed
+
+- **Default model constant replaces hardcoded IDs** — 25+ hardcoded `claude-sonnet-4-20250514` references across routing, providers, cost-tracker, and CLI code now flow through a central `DEFAULT_SONNET_MODEL` constant in `model-registry.ts`. Tier 2 and Tier 3 both route to Sonnet 4.6; Opus 4.7 is reserved as an opt-in escalation target via `MultiModelExecutor`.
+- **Post-audit sweep of lingering model references** — CI workflow, consensus provider, MCP sampling handler, 12 skill evals, 9 test files, opencode plugin, scripts, `.kiro` config, and command docs all updated to the new model IDs.
+- **ADR-092 advisor default model upgraded to Opus 4.7** — The vendor-agnostic advisor now targets Opus 4.7 by default for escalation paths.
+
+### Fixed
+
+- **No fleet breakage at 2026-06-15 Sonnet 4 retirement** — A single-commit model sweep removes every live routing reference to the retiring `claude-sonnet-4-20250514` ID. Remaining references are intentional compatibility metadata (deprecation dates, Bedrock mappings, historical cost entries).
+- **User-facing model listings stay in sync** — `DEFAULT_MODEL_MAPPINGS` and related surface areas now reflect the new model lineup so downstream consumers don't see stale IDs.
+
 ## [3.9.12] - 2026-04-16
 
 ### Fixed

package/assets/agents/v3/qe-coverage-specialist.md

@@ -11,7 +11,7 @@ domain: coverage-analysis
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-fleet-commander.md

@@ -8,7 +8,7 @@ domain: cross-domain
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-pentest-validator.md

@@ -1,14 +1,16 @@
 ---
 name: qe-pentest-validator
 version: "3.6.0"
-updated: "2026-04-12"
+updated: "2026-04-17"
 description: Graduated exploit validation with parallel vulnerability pipelines, browser-based attack execution, and "No Exploit, No Report" quality gate
 v2_compat: null
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 advisor:
   enabled: true
   provider: claude
-  model: claude-opus-4-6
+  model: claude-opus-4-7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-queen-coordinator.md

@@ -14,7 +14,7 @@ dependencies:
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-risk-assessor.md

@@ -8,7 +8,7 @@ v3_new: true
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-root-cause-analyzer.md

@@ -19,7 +19,7 @@ dependencies:
 advisor:
   enabled: true
   provider: openrouter
-  model: anthropic/claude-opus-4
+  model: anthropic/claude-opus-4.7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-security-auditor.md

@@ -1,14 +1,16 @@
 ---
 name: qe-security-auditor
 version: "3.0.0"
-updated: "2026-04-12"
+updated: "2026-04-17"
 description: Security audit specialist with OWASP coverage, compliance validation, and remediation workflows
 v2_compat: null # New in v3
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 advisor:
   enabled: true
   provider: claude
-  model: claude-opus-4-6
+  model: claude-opus-4-7
   max_uses: 3
   redact: strict
 ---

package/assets/agents/v3/qe-security-scanner.md

@@ -1,10 +1,12 @@
 ---
 name: qe-security-scanner
 version: "3.0.0"
-updated: "2026-01-10"
+updated: "2026-04-17"
 description: Comprehensive security scanning with SAST, DAST, dependency scanning, and secrets detection
 v2_compat: qe-security-scanner
 domain: security-compliance
+# ADR-093: security agents default to max effort for highest-stakes reasoning
+effort: max
 dependencies:
   agents:
     - name: qe-dependency-mapper

package/assets/agents/v3/qe-test-architect.md

@@ -11,7 +11,7 @@ domain: test-generation
 advisor:
   enabled: true                    # ADR-092 Phase 0 target agent
   provider: openrouter             # Phase 0 default; any HybridRouter provider supported
-  model: anthropic/claude-opus-4   # Strong reasoning model via OpenRouter
+  model: anthropic/claude-opus-4.7   # Strong reasoning model via OpenRouter
   max_uses: 3                      # Per-task advisor call cap
   budget_usd_per_task: 0.05        # Per-call budget ceiling
   required: false                  # Phase 5 flips this to true for quality-gate enforcement

package/assets/skills/.validation/schemas/skill-eval.schema.json

@@ -26,6 +26,9 @@
       "items": {
         "type": "string",
         "enum": [
+          "claude-opus-4-7",
+          "claude-sonnet-4-6",
+          "claude-haiku-4-5",
           "claude-opus-4-5",
           "claude-sonnet-4",
           "claude-3.5-sonnet",

package/assets/skills/.validation/test-data/sample-output.json

@@ -60,7 +60,7 @@
     "executionTimeMs": 1250,
     "toolsUsed": ["eslint", "typescript", "custom-analyzer"],
     "agentId": "qe-test-architect",
-    "modelUsed": "claude-sonnet-4",
+    "modelUsed": "claude-sonnet-4-6",
     "inputHash": "sha256:abc123def456",
     "parentTaskId": "task-789"
   },

package/assets/skills/a11y-ally/SKILL.md

@@ -1230,7 +1230,7 @@ const client = new Anthropic();
 
 const imageData = fs.readFileSync('/tmp/a11y-work/frames/frame_01.jpg').toString('base64');
 const response = await client.messages.create({
-  model: 'claude-sonnet-4-20250514',
+  model: 'claude-sonnet-4-6',
   max_tokens: 500,
   messages: [{
     role: 'user',

package/assets/skills/accessibility-testing/evals/accessibility-testing.yaml

@@ -28,7 +28,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (ensure minimum quality)
+  - claude-haiku-4-5          # Fast model (ensure minimum quality)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/api-testing-patterns/evals/api-testing-patterns.yaml

@@ -30,8 +30,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4       # Primary model (high accuracy expected)
-  - claude-3-haiku        # Fast model (minimum quality bar)
+  - claude-sonnet-4-6        # Primary model (high accuracy expected)
+  - claude-haiku-4-5          # Fast model (minimum quality bar)
   - gpt-4o                # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/chaos-engineering-resilience/evals/chaos-engineering-resilience.yaml

@@ -35,7 +35,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/contract-testing/evals/contract-testing.yaml

@@ -30,8 +30,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4       # Primary model (high accuracy expected)
-  - claude-3-haiku        # Fast model (minimum quality bar)
+  - claude-sonnet-4-6        # Primary model (high accuracy expected)
+  - claude-haiku-4-5          # Fast model (minimum quality bar)
   - gpt-4o                # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/database-testing/evals/database-testing.yaml

@@ -33,8 +33,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4     # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-sonnet-4-6      # Primary model (high accuracy expected)
+  - claude-haiku-4-5         # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/localization-testing/evals/localization-testing.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
 
 # =============================================================================
 # MCP Integration Configuration

package/assets/skills/n8n-trigger-testing-strategies/evals/n8n-trigger-testing-strategies.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet
-  - claude-3-haiku
+  - claude-haiku-4-5
 
 # =============================================================================
 # MCP Integration Configuration

package/assets/skills/performance-testing/evals/performance-testing.yaml

@@ -35,7 +35,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet    # Primary model (high accuracy expected)
-  - claude-3-haiku       # Fast model (minimum quality threshold)
+  - claude-haiku-4-5          # Fast model (minimum quality threshold)
   - gpt-4o               # Cross-vendor validation
 
 # =============================================================================

package/assets/skills/quality-metrics/evals/quality-metrics.yaml

@@ -32,7 +32,7 @@ description: >
 
 models_to_test:
   - claude-3.5-sonnet
-  - claude-3-haiku
+  - claude-haiku-4-5
 
 # =============================================================================
 # MCP Integration Configuration

package/assets/skills/skills-manifest.json

@@ -939,7 +939,7 @@
   },
   "metadata": {
     "generatedBy": "Agentic QE Fleet",
-    "fleetVersion": "3.9.12",
+    "fleetVersion": "3.9.14",
     "manifestVersion": "1.4.0",
     "lastUpdated": "2026-04-13T00:00:00.000Z",
     "contributors": [

package/assets/skills/testability-scoring/evals/testability-scoring.yaml

@@ -29,8 +29,8 @@ description: >
 # =============================================================================
 
 models_to_test:
-  - claude-sonnet-4-20250514    # Primary model (high accuracy expected)
-  - claude-3-haiku              # Fast model (minimum quality threshold)
+  - claude-sonnet-4-6    # Primary model (high accuracy expected)
+  - claude-haiku-4-5                # Fast model (minimum quality threshold)
   - gpt-4o                      # Cross-vendor validation
 
 # =============================================================================