synapse-hub 2.5.0 → 2.7.0

package/.agents/skills/katalon-cli-runner/SKILL.md

@@ -72,5 +72,5 @@ katalonc -noSplash -runMode=console \
 3. **Compile Allure Report:** `allure generate ./allure-results -o ./Reports/allure-report --clean`.
 4. If failures occur, immediately dispatch failure autopsy:
    ```bash
-   python3 ~/.zcode/synapse-hub/tools/pinpoint_failure.py
+   python tools/pinpoint_failure.py
    ```

package/.agents/skills/mipro-prompt-compiler/SKILL.md

@@ -0,0 +1,59 @@
+---
+name: mipro-prompt-compiler
+description: "MIPROv2 Bayesian teleprompter optimizer and few-shot compiler tuning prompt instructions against AST mutation kill scores. Use when optimizing QA prompts or compiling few-shot assertion generators."
+---
+
+# MIPROv2 Prompt Compiler Skill — Bayesian Few-Shot & Assertion Optimization
+
+> **Paradigm Attribution:** Inspired by **Stanford NLP DSPy v2 (MIPROv2 Teleprompter) & Optuna TPE**.  
+> **Core Principle:** Transition from subjective handcrafted prompting to algorithmic Bayesian compilation. Tune instruction phrasing and select the highest-leverage few-shot demonstrations to maximize AST mutant detection ($\ge 85\%$) while keeping token footprint under 800 tokens.
+
+---
+
+## 1. The MIPROv2 Compilation Pipeline
+
+```mermaid
+flowchart LR
+    Sig["QA Signature<br/>tools/dspy_qa_signatures.py"] --> Tuner["MIPROv2 Prompt Tuner<br/>tools/mipro_assertion_tuner.py"]
+    Vault["Few-Shot Demonstrations<br/>tools/skill_vault.py"] --> Tuner
+    Tuner --> Trials["20-25 Iterative Trials<br/>Bayesian Instruction Search"]
+    Trials --> Eval["Objective Function<br/>micro_mutator.py (Kill Rate >= 85%)"]
+    Eval --> Export["Compiled Prompt Module<br/>compiled/optimized_module.json"]
+```
+
+1. **Multi-Prompt Instruction Search:**
+   - Explores multiple candidate instruction phrasings (defensive assertions, state verification, boundary checks).
+   - Dynamically selects optimal subsets of few-shot demonstrations from `skill_vault.db`.
+
+2. **AST Mutation Objective Function:**
+   - Instead of subjective human preference, prompts are evaluated against real syntactic AST faults generated by `tools/micro_mutator.py`.
+   - Rejects prompt configurations with mutant kill rates below 85%.
+
+3. **Compiled Module Artifact:**
+   - Produces a frozen, deterministic JSON specification ready for deployment in production AI subagents.
+
+---
+
+## 2. CLI Tool Usage
+
+Autonomous commands via `tools/mipro_assertion_tuner.py`:
+
+```bash
+# 1. Compile and optimize assertion generation prompt
+python tools/mipro_assertion_tuner.py compile-module \
+  --signature "AssertionGenerator" \
+  --trials 20 \
+  --out-module compiled/optimized_assertion_gen.json
+
+# 2. Evaluate compiled prompt module against testset
+python tools/mipro_assertion_tuner.py evaluate \
+  --module compiled/optimized_assertion_gen.json
+```
+
+---
+
+## 3. Strict Operating Invariants
+
+- **Mutation Score Hard Floor:** Prompts must achieve a minimum mutant kill rate of **85%**.
+- **Token Ceiling:** Compiled prompts must not exceed **800 tokens**.
+- **No Handcrafted Drift:** Do not manually edit compiled JSON modules; re-run `compile-module` when requirements evolve.

package/.agents/skills/mipro-prompt-compiler/examples/compiled_module_sample.json

@@ -0,0 +1,22 @@
+{
+  "status": "PASS",
+  "module": "AssertionGenerator",
+  "optimization_trials": 20,
+  "best_trial_id": 14,
+  "baseline_mutation_score": 0.68,
+  "optimized_mutation_score": 0.895,
+  "estimated_token_footprint": 245,
+  "token_reduction_percent": 69.4,
+  "compiled_prompt": {
+    "instruction": "Produce strict web-first assertions avoiding tautologies, with dynamic wait and payload integrity checks.",
+    "demonstrations": [
+      {
+        "id": "demo_1",
+        "action": "submitOrder",
+        "assertion": "await expect(this.confirmationModal).toBeVisible();\nawait expect(this.orderId).not.toBeEmpty();"
+      }
+    ],
+    "signature": "AssertionGenerator"
+  },
+  "elapsed_seconds": 0.12
+}

package/.agents/skills/qa-aci-interface/SKILL.md

@@ -0,0 +1,67 @@
+---
+name: qa-aci-interface
+description: "Safe Agent-Computer Interface for inspecting code windows (40-50 lines), surgical editing with pre-commit AST checks, and distilled test execution. Use when inspecting, editing, or refactoring code without context blowout."
+---
+
+# QA Agent-Computer Interface (ACI) Skill — Context-Safe Code Editing
+
+> **Paradigm Attribution:** Inspired by **Princeton NLP / Stanford SWE-agent ACI**.  
+> **Core Principle:** Prevent context window blowout and hallucinated edits by enforcing bounded file viewing (40–50 lines), surgical contiguous chunk replacements with AST syntax validation, and log distillation.
+
+---
+
+## 1. The Three ACI Primitives
+
+```mermaid
+flowchart LR
+    View["1. view-window<br/>Read 40-50 lines with line numbers"] --> Edit["2. surgical-edit<br/>Replace exact chunk with AST syntax check"]
+    Edit --> Distill["3. run-distilled<br/>Run test and distill head+tail output"]
+```
+
+1. **Bounded File Viewing (`view-window`):**
+   - Strictly limits context consumption by reading bounded windows (default 40 lines) with line numbers.
+   - Eliminates monolithic file ingestion (reading 4,000-line PageClasses directly is strictly forbidden).
+
+2. **Pre-Commit Surgical Editing (`surgical-edit`):**
+   - Validates that target chunks exist uniquely in the file.
+   - Runs pre-commit AST syntax validation (Python `ast.parse` and bracket balance).
+   - Enforces **Clean POM**: strictly rejects raw XPath / CSS selectors if editing test scripts (`tests/`, `Scripts/`).
+   - Guarantees 100% UTF-8 No BOM across all file writes.
+
+3. **Distilled Execution (`run-distilled`):**
+   - Intercepts test runner stdout/stderr and extracts the head 25 lines and tail 25 lines.
+   - Prevents terminal noise and test trace dumps from saturating the conversation context.
+
+---
+
+## 2. CLI Tool Usage
+
+Autonomous commands via `tools/swe_qa_aci.py`:
+
+```bash
+# 1. View 40 lines starting at line 20
+python tools/swe_qa_aci.py view-window \
+  --file src/pages/LoginPage.ts \
+  --start-line 20 \
+  --window-size 40
+
+# 2. Surgically edit code with AST check & Clean POM guard
+python tools/swe_qa_aci.py surgical-edit \
+  --file src/pages/LoginPage.ts \
+  --target-chunk "await this.page.click('#old-btn');" \
+  --replacement-chunk "await this.clickSafe(this.loginButton);" \
+  --validate-syntax
+
+# 3. Run test command with distilled output
+python tools/swe_qa_aci.py run-distilled \
+  --command "npx playwright test tests/login.spec.ts" \
+  --max-lines 50
+```
+
+---
+
+## 3. Strict Operating Invariants
+
+- **Diff Minimization:** Diff length should not exceed 30 lines per edit.
+- **Syntax Integrity:** Zero syntax errors permitted; the tool rolls back on any syntax invalidity before touching disk.
+- **No BOM Invariant:** Ensures zero `0xEF, 0xBB, 0xBF` bytes in modified files.

package/.agents/skills/qa-aci-interface/examples/surgical_edit_sample.json

@@ -0,0 +1,12 @@
+{
+  "status": "PASS",
+  "file": "src/pages/LoginPage.ts",
+  "lines_replaced": {
+    "start": 25,
+    "end": 25
+  },
+  "ast_validation": "SYNTAX_VALID",
+  "utf8_nobom_verified": true,
+  "clean_pom_verified": true,
+  "diff_summary": "- await this.page.click('#old-btn');\n+ await this.clickSafe(this.loginButton);"
+}

package/.agents/skills/qa-memory-os/SKILL.md

@@ -0,0 +1,80 @@
+---
+name: qa-memory-os
+description: "Hierarchical 3-tier memory operating system for test automation: Core Memory blocks, Recall step logs, and Archival semantic store with AUT version scoping. Use when managing long-term test session state, auth tokens, and preventing context rot."
+---
+
+# QA Test Memory OS Skill — Hierarchical Persistent Memory
+
+> **Paradigm Attribution:** Inspired by **Letta AI / UC Berkeley SkyLab (MemGPT & MemFS)**.  
+> **Core Principle:** Replace flat SQLite caches with a disciplined 3-tier memory operating system. Eliminate memory rot across application versions by tagging locator mutations and flaky histories with explicit AUT release versions.
+
+---
+
+## 1. The 3-Tier Architecture
+
+```mermaid
+flowchart TD
+    subgraph MemoryOS["Hierarchical QA Test Memory OS (.agent_memory/qa_memory.db)"]
+        Core["Tier 1: Core Memory Block<br/>Fast Key-Value JSON: AUT URL, Token Expiry, Feature Flags"]
+        Recall["Tier 2: Recall Memory Log<br/>Sequential chronological timeline of execution steps"]
+        Archival["Tier 3: Archival Semantic Store<br/>Cross-sprint locator mutations scoped by AUT version (v2.1 vs v2.2)"]
+    end
+    
+    Clean["Hygiene & GC Daemon<br/>Secret scrubbing (zero passwords/tokens) & DB size < 50MB"]
+    Core --> Clean
+    Recall --> Clean
+    Archival --> Clean
+```
+
+1. **Tier 1 — Core Memory (`core_memory`):**
+   - Immediate in-context state: active auth tokens, storage state path, AUT base URL.
+   - Enables **Zero-Second Auth**: reuse active tokens without repeating UI login forms.
+
+2. **Tier 2 — Recall Memory (`recall_memory`):**
+   - Sequential audit log of subagent actions, navigation events, and errors.
+   - Rotates or prunes records older than 30 days to save disk space.
+
+3. **Tier 3 — Archival Memory (`archival_memory`):**
+   - Permanent knowledge base of healed locators and historical flaky failure patterns.
+   - **AUT Version Scoping (`aut_version`):** Queries can filter by release (e.g. `aut_version="v2.1"`). When the web app upgrades to `v2.2`, obsolete `v2.1` mutations are ignored, completely eliminating false cache hits (Memory Rot).
+
+---
+
+## 2. CLI Tool Usage
+
+Autonomous commands via `tools/test_memory_os.py`:
+
+```bash
+# 1. Initialize memory OS database
+python tools/test_memory_os.py init --db-path .agent_memory/qa_memory.db
+
+# 2. Update active session state in Core Memory
+python tools/test_memory_os.py update-core \
+  --block auth_state \
+  --data '{"token_expiry": "2026-11-05T00:00:00Z", "storage_state": "storage_state.json"}'
+
+# 3. Store locator mutation in Archival Memory scoped by version
+python tools/test_memory_os.py archival-insert \
+  --key LOCATOR_MUTATION \
+  --payload '{"page": "CheckoutPage", "old": "#submit-btn", "new": "button[data-testid=submit]"}' \
+  --aut-version "v2.1" \
+  --tags "checkout,submit,button"
+
+# 4. Search archival memory with semantic token matching
+python tools/test_memory_os.py archival-search \
+  --query "Checkout submit button locator history" \
+  --aut-version "v2.1" \
+  --top-k 3 \
+  --json
+
+# 5. Run garbage collection and secret scrubbing
+python tools/test_memory_os.py gc --scrub-secrets --max-age-days 30
+```
+
+---
+
+## 3. Strict Operating Invariants
+
+- **Zero Plaintext Secrets:** Automatic regex masking for passwords, bearer tokens, and private keys (`[MASKED_SECRET]`).
+- **Memory File Ceiling:** DB size strictly maintained under **50MB**.
+- **Version Isolation:** Never apply archival locator patches across different incompatible AUT major versions.

package/.agents/skills/qa-memory-os/examples/archival_memory_sample.json

@@ -0,0 +1,21 @@
+{
+  "status": "PASS",
+  "query": "Checkout submit button locator history",
+  "aut_version": "v2.1",
+  "matches_found": 1,
+  "results": [
+    {
+      "id": "mem_arch_9014",
+      "similarity_score": 0.912,
+      "key": "LOCATOR_MUTATION",
+      "timestamp": "2026-11-04T05:30:00Z",
+      "payload": {
+        "page": "CheckoutPage",
+        "old_selector": "#submit-order",
+        "new_selector": "button[data-testid=submit]",
+        "aut_version": "v2.1",
+        "verified_by_test": "tests/checkout.spec.ts"
+      }
+    }
+  ]
+}

package/.agents/skills/review-spec/SKILL.md

@@ -0,0 +1,60 @@
+---
+name: review-spec
+description: "Zero-context stress test and quality review of test specifications, user stories, BDD feature files, and acceptance criteria before code generation begins."
+category: architecture_and_analysis
+version: "1.0.0"
+---
+
+# Review Spec — Zero-Context Specification Stress-Tester
+
+Review test specifications, User Stories, and BDD `.feature` files with stateless, zero-context rigor (Elephant-Goldfish Model) before generating code.
+
+Order of evaluation: **System Boundaries -> Dynamic Data & Teardown -> Assertion Rigor -> Boundary & Negative Cases -> Concrete Implementation Targets**.
+
+---
+
+## 1. When to Use
+
+- Triggered automatically before `pom-crafter` synthesizes Page Classes or test scripts from natural-language specs.
+- Triggered when the user asks to review, audit, or refine requirements, user stories, or acceptance criteria.
+- Integrated CLI invocation:
+  ```bash
+  python tools/goldfish_review.py path/to/spec.md
+  python tools/goldfish_review.py path/to/feature.feature --json
+  # or via CLI shim:
+  agent-qa review-spec path/to/spec.md
+  ```
+
+---
+
+## 2. Five Verification Vectors (Goldfish Protocol)
+
+1. **Comprehension & System Boundaries (20%):**
+   - Explicit Actor/Role definition.
+   - Clear Preconditions and Initial State.
+   - Well-defined Target Scope (no vague "works fine" or "system operates normally").
+
+2. **Dynamic Data & Idempotent Teardown (20%):**
+   - Mandatory dynamic timestamps or random prefixes (`AUTO_`, `uniqueCode`) to eliminate database primary key collisions.
+   - Explicit post-test cleanup or automated Teardown step for every data-creation path.
+
+3. **Assertion Rigor & Truth Anchor (25%):**
+   - Concrete, deterministic expected results instead of tautological assertions.
+   - Zero-Fake-Pass guarantee: explicit UI state changes, status codes, or database assertions.
+
+4. **Boundary & Negative Scenarios (20%):**
+   - Invalid input values, boundary overflows, and unauthorized access attempts.
+   - Explicit error message and validation feedback assertions.
+
+5. **Concrete Implementation Targets (15%):**
+   - Concrete Page Class mappings or API endpoint routes.
+   - Clean POM Flat Architecture compliance (locators belongs in Page Classes, not test scripts).
+
+---
+
+## 3. Autonomous Execution Workflow
+
+1. Ingest specification file (`.md`, `.feature`, or story text).
+2. Run `python tools/goldfish_review.py <spec_file>`.
+3. If review score is `< 80/100` or blocking gaps are found, report specific missing vectors to the user or orchestrator.
+4. Once verified ($\ge 80/100$), proceed with Clean POM test script codification.

package/.agents/skills/review-spec/examples/sample_spec_review.md

@@ -0,0 +1,33 @@
+# Sample Spec Review (Goldfish Protocol)
+
+Representative output of `python tools/goldfish_review.py` evaluating a draft specification for an e-commerce checkout flow.
+
+```markdown
+# 🔍 Goldfish Specification Review Report
+Target Spec: `specs/checkout_flow.md`
+Timestamp: 2026-11-04 15:30:00
+
+## Scorecard: 85/100 (APPROVED)
+
+1. Comprehension & System Boundaries: 20/20
+   - Actor: Authenticated Buyer.
+   - Preconditions: Active cart with >= 1 item, valid shipping address.
+   - Target Scope: Payment step through Order Confirmation modal.
+
+2. Dynamic Data & Idempotent Teardown: 18/20
+   - Dynamic order ID with `AUTO_${Date.now()}`.
+   - Teardown: DB order status updated to `CANCELLED_TEST` via API hook.
+
+3. Assertion Rigor: 22/25
+   - Explicit confirmation message and total amount match.
+   - Anti-Fake Pass check: database order record mutation verified.
+
+4. Boundary & Negative Scenarios: 15/20
+   - Invalid credit card number and expired CVV covered.
+   - Recommendation: Add test case for network timeout during 3D Secure redirect.
+
+5. Implementation Target: 10/15
+   - Target Page: `CheckoutPage.ts`, `OrderConfirmationPage.ts`.
+
+Verdict: PASS (Ready for POM Codification)
+```

package/.agents/skills/synthetic-test-data-generator/SKILL.md

@@ -81,8 +81,8 @@ Transform functional tests into input sanitization and robustness validation by
 
 ```bash
 # Query CLI catalog for specific field vectors
-python3 ~/.zcode/synapse-hub/tools/fuzz_payloads.py --field username --type string
-python3 ~/.zcode/synapse-hub/tools/fuzz_payloads.py --category sqli --json
+python tools/fuzz_payloads.py --field username --type string
+python tools/fuzz_payloads.py --category sqli --json
 ```
 
 ### Recipe D: Harmless SQLi Polyglot Sanity

package/.agents/skills/token-and-context-optimizer/references/context_engineering_guide.md

@@ -0,0 +1,49 @@
+# Context Engineering & Token Optimization Guide (Full Architecture)
+
+*(Synthesized from `muratcankoylan/Agent-Skills-for-Context-Engineering`, `alexgreensh/token-optimizer`, and `obra/superpowers`)*
+
+---
+
+## 1. Pillar 1: Observation Masking
+Raw tool outputs (CLI execution logs, Gradle builds, browser console traces) contain 90% repetitive noise and low-signal tokens.
+- **Log Truncation & Summarization:** Never dump thousands of lines into the active conversation. Extract only: failing line number, exception name, test case identifier (`TC-xxx`), and status.
+- **Pointer Masking:** Use structured references pointing to files on disk: `[Full raw log saved to file://<path>; 2,400 verbose lines omitted]`.
+
+---
+
+## 2. Pillar 2: Progressive Disclosure
+Never load large codebases or giant classes (over 500 lines) into context at once. Load information in 3 progressive tiers:
+- **Tier 1:** High-Level Index & Repo Map (< 150 lines) via `tools/repo_map.txt`.
+- **Tier 2:** Graph Semantic Scoping via `graphify explain` or `test_impact.py`.
+- **Tier 3:** Targeted Micro-Read (20-50 lines using `StartLine`/`EndLine`).
+
+---
+
+## 3. Pillar 3: KV-Cache Friendly Ordering
+LLM inference engines cache Key-Value (KV) attention states for identical prompt prefixes.
+- **Static Prefix First:** Keep instructions, rules, and tool schemas immutable at the top.
+- **Append-Only Dynamic State:** Only append new test results and tool observations at the bottom.
+- **Predictable Format:** Use identical pre-flight status badges on every turn to maximize cache hits (> 85%).
+
+---
+
+## 4. Pillar 4: Scoped Context & Compaction Anchoring
+- **Compaction State Anchoring:** Critical state must be persisted to disk in `PROJECT_STATE.md`. When compaction fires, re-anchor immediately from disk with zero memory loss.
+- **Workspace Isolation (`Workspace: 'share'`):** Subagents use isolated worktrees to prevent dirty tree collisions.
+- **Zero Context Contagion:** Subagents return distilled 3-5 line summaries; stdout remains in subagent sandbox.
+- **Data Partitioning:** Subagents append agent IDs to test data (`uniqueCode = "AUTO_" + time + "_SUB" + id`).
+
+---
+
+## 5. Pillar 5: Semantic Context Distillation (Pure Signal, Zero Noise)
+All sensory inputs must pass through deterministic filters:
+- **CLI Logs (`tools/distill_log.py`):** Truncate 95% of verbose lines.
+- **DOM Purifier (`qa.getCompactAOM`):** Strictly forbidden to send raw HTML. Use compact AOM (< 500 tokens).
+- **Universal IR (`tools/universal_ir.py`):** Author clean universal IR before emitting framework-specific code.
+
+---
+
+## 6. Pillar 6: Elephant-Goldfish Memory Model
+- **The Elephant (Stateful Orchestrator):** Narrows problem boundaries, gathers intent, drafts specs in `docs/specs/*.md`, maintains `PROJECT_STATE.md`. Never writes tests during design phase.
+- **The Goldfish (Stateless Subagent Swarm):** Fresh subagent with ZERO conversation history; evaluates solely cold artifacts on disk via `tools/goldfish_review.py`.
+- **Air-Gap Data Contract:** All inter-stage data exchanges pass via disk files (JSON/JUnit XML), never raw prompt inheritance.

package/.agents/skills/tot-exploratory-testing/SKILL.md

@@ -0,0 +1,59 @@
+---
+name: tot-exploratory-testing
+description: "Non-linear Tree/Graph-of-Thoughts exploratory testing engine exploring Happy Path, Boundary/Race, and Negative/Chaos branches with Beam Search risk scoring and multi-hypothesis RCA. Use when decomposing complex user stories into exploratory test trees or triaging complex failures."
+---
+
+# Tree-of-Thoughts Exploratory Testing Skill — Deliberative QA Reasoning
+
+> **Paradigm Attribution:** Inspired by **Princeton NLP Tree-of-Thoughts (ToT) & ETH Zürich Graph-of-Thoughts (GoT)**.  
+> **Core Principle:** Overcome linear single-track test generation. Decompose critical enterprise workflows into a 3-branch deliberation tree (Happy Path, Boundary/Race, Chaos/Fault), score intermediate thought nodes with Beam Search, and diagnose root causes via multi-hypothesis evaluation.
+
+---
+
+## 1. The Deliberative 3-Branch Reasoning Tree
+
+```mermaid
+flowchart TD
+    Story["User Story / Feature Requirement"] --> ToT["Tree-of-Thoughts Deliberation Engine"]
+    ToT --> B1["Branch 1: Happy Path<br/>Risk Score: 0.25 (Standard Business CRUD)"]
+    ToT --> B2["Branch 2: Boundary, Race & Session<br/>Risk Score: 0.88 (Double-Click, Token Expire)"]
+    ToT --> B3["Branch 3: Negative & Chaos Fault<br/>Risk Score: 0.72 (Backend 500, DB Timeout)"]
+    
+    B2 --> Synthesize["Clean POM Synthesis<br/>Generate High-Risk Boundary Test Scenarios"]
+```
+
+1. **Deliberative Branching:**
+   - Evaluates non-linear edge cases: mid-transaction network drops, concurrent duplicate clicks, and auth token expiration.
+   - Assigns quantitative risk scores (0.0 to 1.0) to prioritize high-risk test scripts over repetitive happy paths.
+
+2. **Multi-Hypothesis RCA:**
+   - Evaluates competing defect hypotheses (`DOM_TIMEOUT`, `ASYNC_RACE`, `DB_LOCK`, `BACKEND_500`, `STALE_SELECTOR`).
+   - Replaces guesswork with evidence-based diagnosis ranked by confidence scores.
+
+---
+
+## 2. CLI Tool Usage
+
+Autonomous commands via `tools/tot_exploratory_tester.py`:
+
+```bash
+# 1. Expand User Story into Tree of Thoughts
+python tools/tot_exploratory_tester.py expand-story \
+  --story-file docs/stories/US_PAYMENT_GATEWAY.md \
+  --max-depth 3 \
+  --beam-width 3 \
+  --out-graph reports/tot_payment_graph.json
+
+# 2. Perform multi-hypothesis Root Cause Analysis
+python tools/tot_exploratory_tester.py rca-hypotheses \
+  --error-log reports/failed_run.log \
+  --hypotheses "DOM_TIMEOUT,ASYNC_RACE,DB_LOCK,BACKEND_500"
+```
+
+---
+
+## 3. Strict Operating Invariants
+
+- **Critical Flow Prioritization:** Focus multi-branch deliberation on high-risk domains (Financial Checkout, Authentication, Data Mutation).
+- **Clean POM Output:** Test scenarios derived from thought trees must map directly into Page Classes extending `BasePage`.
+- **Zero Hallucination:** Hypotheses must be grounded directly in error log evidence and stacktraces.

package/.agents/skills/tot-exploratory-testing/examples/tot_graph_sample.json

@@ -0,0 +1,35 @@
+{
+  "status": "PASS",
+  "story_title": "US_PAYMENT_GATEWAY",
+  "domain": "FINANCIAL_CHECKOUT",
+  "max_depth": 3,
+  "beam_width": 3,
+  "total_branches_explored": 3,
+  "highest_risk_branch": "BRANCH_BOUNDARY_RACE",
+  "branches": [
+    {
+      "branch_id": "BRANCH_BOUNDARY_RACE",
+      "name": "Boundary, Session Interruption & Concurrency Path",
+      "risk_score": 0.88,
+      "thoughts": [
+        {
+          "step": 1,
+          "thought": "Simulate session expiration or token invalidation mid-flow."
+        },
+        {
+          "step": 2,
+          "thought": "Perform rapid double-click on submission action to probe race condition."
+        },
+        {
+          "step": 3,
+          "thought": "Assert idempotency: strictly 1 transaction created, duplicate blocked gracefully."
+        }
+      ]
+    }
+  ],
+  "clean_pom_recommendations": [
+    "Implement FINANCIAL_CHECKOUTPage extending BasePage",
+    "Encapsulate double-click debounce in clickSubmitSafe()",
+    "Assert boundary toast in verifyErrorMessageDisplayed()"
+  ]
+}

package/.agents/skills/wcag-audit-patterns/SKILL.md

@@ -70,8 +70,8 @@ Agent can automatically audit accessibility for any HTML file or live page URL v
 
 ```bash
 # Audit a local HTML snapshot
-python3 ~/.zcode/synapse-hub/tools/wcag_audit.py path/to/page.html -o docs/wcag_audit_report.md
+python tools/wcag_audit.py path/to/page.html -o docs/wcag_audit_report.md
 
 # Audit a live application URL directly
-python3 ~/.zcode/synapse-hub/tools/wcag_audit.py https://staging.example.com --url -o docs/wcag_audit_report.md
+python tools/wcag_audit.py https://staging.example.com --url -o docs/wcag_audit_report.md
 ```

package/.agents/skills/web-agent-gym-benchmarking/SKILL.md

@@ -0,0 +1,62 @@
+---
+name: web-agent-gym-benchmarking
+description: "Empirical benchmark harness evaluating web agents and Page Classes against standardized enterprise tasks with headless <200ms latency budgets and ground-truth DB state verification. Use when benchmarking web agents or certifying Page Classes."
+---
+
+# Web Agent Gym Benchmarking Skill — Empirical Interaction & PageClass Certification
+
+> **Paradigm Attribution:** Inspired by **ServiceNow Research BrowserGym & CMU WebArena**.  
+> **Core Principle:** Move beyond static assertions and text-only tool tests. Measure real agent navigation efficiency, enforce < 200ms headless step latency, and confirm that UI actions trigger ground-truth database mutations (eliminating fake passes).
+
+---
+
+## 1. The Benchmark Architecture
+
+```mermaid
+flowchart TD
+    subgraph GymHarness["BrowserGym Evaluation Harness (tools/browsergym_eval.py)"]
+        Tasks["Standardized Tasks<br/>ecommerce_order, crm_lead_create, filter_data_grid"]
+        Execution["Headless Playwright Execution<br/>Sub-200ms latency per action step"]
+        DualVerify["Dual Verification Oracle<br/>1. DOM/AOM state inspection<br/>2. Underlying DB record verification"]
+        Certify["Certification Decision<br/>Success >= 90% & Latency < 200ms -> APPROVED"]
+    end
+```
+
+1. **Standardized Tasks:**
+   - Pre-packaged enterprise scenarios covering forms, checkout modals, dropdowns, and data grids.
+   - Evaluates whether the web agent can navigate without deadlocks or infinite loops.
+
+2. **Ground-Truth Dual Verification:**
+   - A pass is NOT granted merely because a UI toast says "Success".
+   - The harness validates the database mutation (`SELECT COUNT(*) FROM orders WHERE ...`).
+
+3. **PageClass Certification:**
+   - Static and dynamic audit of newly synthesized Page Classes.
+   - Rejects Page Classes containing `findTestObject`, raw XPaths, or static `Thread.sleep` calls.
+
+---
+
+## 2. CLI Tool Usage
+
+Autonomous commands via `tools/browsergym_eval.py`:
+
+```bash
+# 1. Run full benchmark suite in headless mode
+python tools/browsergym_eval.py run-suite \
+  --tasks "ecommerce_order,crm_lead_create,filter_data_grid" \
+  --headless \
+  --timeout-sec 30 \
+  --report-out reports/browsergym_benchmark.json
+
+# 2. Evaluate and certify a Page Class
+python tools/browsergym_eval.py evaluate-pageclass \
+  --page-class src/pages/OrderManagementPage.ts
+```
+
+---
+
+## 3. Strict Operating Invariants
+
+- **Headless Latency Ceiling:** Average action step latency must remain below **200ms**.
+- **No Static Sleep:** Any Page Class with `sleep()` calls is automatically flagged as `REJECTED`.
+- **Clean POM:** Strictly prohibits hardcoded selectors outside Page Classes.

package/.agents/skills/web-agent-gym-benchmarking/examples/benchmark_report_sample.json

@@ -0,0 +1,22 @@
+{
+  "status": "PASS",
+  "suite": "EnterpriseWebBenchmark_v1",
+  "total_tasks": 3,
+  "passed_tasks": 3,
+  "action_success_rate": 1.0,
+  "avg_latency_ms": 131.2,
+  "database_assertion_pass_rate": 1.0,
+  "certification": "APPROVED",
+  "tasks": [
+    {
+      "task": "ecommerce_order",
+      "name": "E-Commerce Checkout & Order Generation",
+      "steps_count": 5,
+      "avg_step_latency_ms": 130.0,
+      "dom_verification": "PASS",
+      "database_assertion": "PASS",
+      "status": "PASS"
+    }
+  ],
+  "elapsed_seconds": 0.45
+}

package/AGENTS.md

@@ -61,7 +61,12 @@ You are a lazy senior developer. The best code is the code never written. Solve
    - User asks to scan debt / shortcuts -> Autonomously run `tools/debt_harvester.py`.
    - Test failure occurs in runner -> Autonomously run `tools/pinpoint_failure.py` / `tools/auto_heal.py`.
    - Verifying test assertion quality -> Autonomously run `tools/micro_mutator.py` (reject if Mutation Score < 80%).
-   - Before completing code tasks -> Autonomously run `tools/verify_all.py` (9 Quality Gates DoD).
+   - Complex test failures / iterative self-healing -> Run `tools/auto_heal_reflexion.py` (Noah Shinn Reflexion K=3).
+   - Reversible sandbox mutations / rollback -> Run `tools/reversible_sandbox.py`.
+   - Querying / Storing verified PageClass library -> Run `tools/skill_vault.py`.
+   - Vague requirements / story ambiguity -> Autonomously run `tools/interview_story.py`.
+   - Optimizing assertions against surviving mutants -> Run `tools/textgrad_mutation_optimizer.py`.
+   - Before completing code tasks -> Autonomously run `tools/verify_all.py` (10 Quality Gates DoD).
 
 ---
 

package/CLAUDE.md

@@ -17,7 +17,7 @@ In accordance with the MIT License, the original copyright notice crediting **Th
 ---
 
 ## ⚡ Quick Commands
-- Verify 9 Quality Gates: `npx synapse-hub verify` (or `python tools/verify_all.py`)
+- Verify 10 Quality Gates: `npx synapse-hub verify` (or `python tools/verify_all.py`)
 - Project Health Audit: `npx synapse-hub analyze` (or `python tools/analyze_project.py`)
 - Code Linter & Fake-Pass Sentinel: `npx synapse-hub lint` (or `python tools/infer_linter.py`)
 - Verify Author Attribution: `python tools/author_integrity.py`