@jakkrichm/create-nexus-devflow 2.0.27 → 2.2.0

package/template/.claude/skills/try/SKILL.md

@@ -1,108 +1,120 @@
 ---
 name: try
-description: "[Devflow] Generate human manual QA review walkthrough guide (where to go, what to click, what to expect)."
+description: "[devflow][B] Generate a human manual try guide for the current or most recently completed Blueprint feature, fix, or rollback. Reads the spec, project commands, and available app context, then tells the user exactly what to start, where to go, what to click or run, what to expect, and what would count as wrong. Read-only. Use when the user runs /try, invokes $try, asks how to test manually, asks where to click, asks how to see the change, or wants a manual review path after /implement, /autopilot, /check, or /complete."
 ---
 
-# try - Manual QA & Review Walkthrough Guide
+# try - manual review guide
 
 Where this sits in the workflow:
 
-```text
-40-execute or 50-verify or 60-report  ->  [try]  ->  human review & verification
-(work implemented / verified)                  (manual    (where to go,
-                                               steps)     what to click)
-```
+    /implement or /complete  ->  [try]  ->  human review
+    (work exists)                (manual   (where to go,
+                                  path)     what to click)
 
-`50-verify` proves behavior from automated QA and test runs. `try` gives the user and testers an actionable, step-by-step manual walkthrough: start this command, open this route, click these controls, expect this result, and watch for these failure signs.
+`/check` proves behavior from the agent side. `/try` gives the user a practical
+manual walkthrough: start this command, open this route, click these controls,
+expect this result, and watch for these failure signs.
 
-It is always **read-only 100%**. It does not edit files, install dependencies, commit, merge, push, or run destructive commands.
+It is always read-only. It does not edit files, install dependencies, commit,
+merge, push, or run destructive commands.
 
 ## Input
 
 Optional scope:
 
-- **no argument**: use the active run in `devflow/context/current-stage.md` (or the latest completed run under `devflow/runs/`)
-- `latest`: use the most recent completed run in `devflow/runs/`
-- a run ID or path: e.g. `try RUN-002-add-onboard-adopt-doctor-skills`
-- a specific route, endpoint, or CLI command: focus the walkthrough on that surface
+- no argument: use the active feature, fix, or rollback in
+  `devflow/context/current-feature.md`
+- `latest`: use the most recent archive under `devflow/history/features/`,
+  `devflow/history/fixes/`, or `devflow/history/rollbacks/`
+- a step name or number: focus the guide on that current-feature step
+- a path, route, or command: include it as the main thing to try
 
-If there is no active run and no run history, ask what change or feature the user wants to test manually.
+If there is no active feature and no useful archive, ask what change the user
+wants to try instead of guessing.
 
-## Step 1 - Find The Work To Explain
+## Step 1 - find the work to explain
 
 Read:
 
-- `AGENTS.md` (Commands section)
-- `devflow/context/current-stage.md`
+- `AGENTS.md`
+- `devflow/context/current-feature.md`
 - `devflow/context/project-overview.md`
 - `devflow/context/coding-standards.md`
-- Active run artifacts: `20-spec.md`, `30-plan.md`, `40-execute.md`, `50-verify.md` (or archived run artifacts)
-- Current git branch and status
+- `devflow/build-plan.md`
+- latest files under `devflow/history/features/`,
+  `devflow/history/fixes/`, and `devflow/history/rollbacks/`, if the current
+  feature is reset
+- git branch and working tree status
 
-Do not dump the full spec. Extract only the concrete routes, screens, CLI commands, API endpoints, test data, and expected outcomes that a human needs to verify the feature.
+Prefer the active spec. If `current-feature.md` is the reset stub, use the most
+recent archived feature, fix, or rollback by filename or modification time and
+say that is what you used.
 
-## Step 2 - Identify How To Run The App
+Do not dump the spec. Pull out the routes, commands, UI surfaces, CLI commands,
+API endpoints, data states, and done-whens that matter for a human trying it.
+For a rollback, lead with the path that proves the removed behavior is gone, then
+include one unaffected regression path from the rollback spec.
 
-Check the Commands section in `AGENTS.md`. Adapt to the project type:
+## Step 2 - identify how to run the app
 
-- **Web app**: dev server command (e.g. `npm run dev`), local URL (e.g. `http://localhost:3000`), and specific route/screen.
-- **Server / API**: start command, base URL, endpoint, HTTP method, payload, and expected response shape.
-- **CLI**: exact CLI command, flags, arguments, and expected terminal output.
-- **Library / Module**: sample usage snippet, REPL call, or interactive test command.
-- **Fullstack / Microservices**: minimal combined startup commands (e.g. backend + frontend).
+Use the Commands section in `AGENTS.md`. Match the project type:
 
-If a startup command is missing or unclear in `AGENTS.md`, report that as a configuration gap rather than guessing.
+- **Web app** - dev server command, URL, and the route or screen to open.
+- **Server/API** - server command, base URL, endpoint, method, and expected
+  response shape.
+- **CLI** - exact command(s), arguments, and expected output.
+- **Library** - example command, test fixture, REPL snippet, or sample call.
+- **Hybrid app** - list the smallest set of commands needed, such as backend plus
+  web dev server.
 
-## Step 3 - Produce The Manual Walkthrough Guide
+If the app may already be running, say how to reuse it. If a command is missing
+from `AGENTS.md`, report that as a gap rather than inventing certainty.
 
-Format the guide with these 5 standard sections:
+## Step 3 - write the manual guide
 
-1. **1. Start** - exact command(s) to start the system and where to run them.
-2. **2. Open** - URL(s), screens, tabs, API endpoints, or terminal locations.
-3. **3. Do** - specific clicks, form inputs, toggles, selections, or arguments.
-4. **4. Expect** - expected visible UI change, response payload, state change, output, or absence of errors.
-5. **5. Watch For** - common failure symptoms, console errors, network 4xx/5xx errors, stale state, layout breakage, or safety warnings.
+Produce a short guide with these sections:
 
-### Example Walkthrough Style:
+1. **Start** - commands to run and where to run them.
+2. **Open** - URLs, screens, tabs, API endpoints, or CLI commands.
+3. **Do** - clicks, inputs, selections, or command arguments.
+4. **Expect** - visible result, output, response, state change, file, or lack of
+   error.
+5. **Watch For** - common wrong outcomes, console or network errors, stale data,
+   missing fields, bad empty states, layout issues, or safety warnings.
 
-```markdown
-### 1. Start
-Run `npm run dev` in the project root.
+Keep it concrete. Prefer:
 
-### 2. Open
-Navigate to `http://localhost:3000/dashboard/reports`.
+    Open http://127.0.0.1:7788/api/snapshot
+    Expect a JSON object with `generated_at`, `services`, `projects`, and
+    `conflicts`.
 
-### 3. Do
-1. Click the **"Export Report"** button in the top right.
-2. Select **"Format: HTML"** from the dropdown.
-3. Click **"Download"**.
+Avoid:
 
-### 4. Expect
-- A new file `report.html` is downloaded.
-- The UI displays a green success toast: *"Report exported successfully"*.
-- The downloaded HTML file opens in the browser with full styling and charts.
+    Check that the snapshot works.
 
-### 5. Watch For
-- Spinner hanging indefinitely without downloading.
-- Console error related to `Blob` or `URL.createObjectURL`.
-- Broken styling or unrendered Mermaid charts in the exported HTML.
-```
+## Step 4 - include confidence and gaps
 
-## Step 4 - Include Confidence, Best Signal & Gaps
+End with:
 
-Conclude with:
+- **Best signal** - the one thing the user should try first.
+- **Optional deeper checks** - only if useful.
+- **Gaps** - anything the guide cannot know from the docs, such as missing route
+  names, seed data, credentials, or external services.
 
-- **Best Signal**: the single most critical action the user should try first to confirm functionality.
-- **Edge Cases & Error Flows**: optional secondary checks (e.g. invalid input, empty state, network failure).
-- **Gaps & Assumptions**: anything the guide cannot know for certain (e.g. seed credentials, third-party API keys, required database fixtures).
+If the feature is not user-visible, say so and provide the closest manual signal,
+such as an API response, CLI output, log line, or unit test command.
 
 ## Rules
 
-- **Strictly Read-Only**: Never edit code, commit, push, install packages, or mutate repository state.
-- **Do Not Run Automatically**: Do not launch the server or run commands unless the user explicitly instructs you to do so in the chat.
-- **Be Concrete & Specific**: Provide exact URLs, button labels, and payloads.
-- **Acknowledge Uncertainty**: If a route or behavior is not specified in the spec, state it clearly as an assumption.
+- Read-only only. Do not edit, commit, merge, push, install, or delete.
+- Do not run the app unless the user explicitly asks you to try it for them.
+- Do not pretend a path is known when the spec does not say it. Give the best
+  likely path and label uncertainty.
+- Keep the guide short enough to follow while the app is open.
+- Match the project's commands from `AGENTS.md`.
 
-## Output Formatting
+## Formatting
 
-Follow the project conventions in `devflow/context/ai-interaction.md`: concise, scannable markdown with numbered steps and clear bold headers.
+Format the output to match the project's conventions in
+`devflow/context/ai-interaction.md`: concise, scannable markdown, with numbered
+steps for the manual path and short bullets for warnings.

package/template/AGENTS.md

@@ -25,8 +25,9 @@ The workflow and skills are exposed through tool-specific adapters:
 
 - **OpenAI Codex, Google Antigravity & GitHub Copilot**: `.agents/skills/<skill>/SKILL.md`
 - **Claude Code**: `.claude/skills/<skill>/SKILL.md`
+- **OpenCode**: `AGENTS.md` plus the compatible `.agents/skills/` or `.claude/skills/` tree already installed for the selected tools
 
-Unused adapter families can be removed. Codex, Antigravity, and GitHub Copilot share `.agents/` and `AGENTS.md`. Claude Code projects keep `.claude/` and `AGENTS.md` (via `CLAUDE.md`).
+Unused adapter families can be removed. Codex, Antigravity, GitHub Copilot, and OpenCode share `.agents/` and `AGENTS.md`. OpenCode can also reuse `.claude/` when Claude Code is selected. Claude Code projects keep `.claude/` and `AGENTS.md` (via `CLAUDE.md`). Do not duplicate the same DevFlow skills under `.opencode/skills/`; OpenCode already discovers the compatible trees.
 
 ### Universal Invocation & Agent Directives:
 

package/template/devflow/build-plan.md

@@ -1,40 +1,70 @@
 # 📋 Build Plan (User-Owned Feature Queue)
 
 > **Document Type**: Build Plan (User-Owned)  
-> **Purpose**: Master sequential feature build queue with dependencies and sizing. Inspected by `/brief` and consumed by `/feature`.
+> **Purpose**: รายการคิวฟีเจอร์การพัฒนาปรับปรุงระบบตามลำดับ พร้อมการประเมินขนาด (Sizing) และความสัมพันธ์ (Dependencies) สำหรับ `/brief` และ `/feature`
 
 ---
 
-## 🚀 Phase 1: Core Foundation & MVP
+## 🚀 Phase 1: Hard Quality Gates & Pre-commit Enforcement
 
-- [ ] **1. Project Baseline & Context Setup** `[Size: S]`
+- [x] **1. Automated Quality Gatekeeper & Pre-commit Hook Integration** `[Size: S]`
   - *Dependencies*: None
-  - *Scope*: Configure project scaffolding, coding standards, and initial workspace verification.
-- [ ] **2. Core Data Models & Store** `[Size: M]`
+  - *Scope*: ผสาน `nexus-devflow check-gate` เข้าสู่ Git Pre-commit Hooks และ CI เพื่อบล็อกการ Commit/Merge เมื่อมี Unchecked Tasks, Test ล้มเหลว หรือมี P0/P1 Finding ค้างอยู่
+
+---
+
+## ⚡ Phase 2: DevFlow Model Context Protocol (MCP) Server Hub
+
+- [x] **2. DevFlow MCP Server Hub & Type-Safe Schema Engine (`IDEA-004`)** `[Size: M]`
   - *Dependencies*: Feature 1
-  - *Scope*: Implement foundational data schemas, types, and persistence layer.
-- [ ] **3. Primary User Interface & Navigation** `[Size: M]`
+  - *Scope*: สร้าง Subcommand `nexus-devflow mcp` เปิด JSON-RPC Server พร้อมเครื่องมือ Typed Tools (`update_task`, `record_finding`, `get_stage_context`) ด้วย Zod Schema Validation
+
+---
+
+## 🌿 Phase 3: Branch-Scoped Context Isolation
+
+- [x] **3. Branch-Scoped Context Isolation & Dynamic Router** `[Size: M]`
   - *Dependencies*: Feature 2
-  - *Scope*: Build core views, interactive components, and responsive layout.
+  - *Scope*: แยกจัดเก็บ State ตาม Git Branch ใน `.nexus/branches/<branch-name>/`, ระบบ Canonical Dynamic Router สำหรับ AI Agent และระบบ Auto-Cleanup หลัง `/complete`
+
+---
+
+## ✂️ Phase 4: JIT Context Slicing & Token Optimizer
+
+- [x] **4. Just-In-Time (JIT) Dynamic Context Slicing Engine** `[Size: M]`
+  - *Dependencies*: Feature 3
+  - *Scope*: สร้างตัวตัดตอนบริบท (Context Slicer) สำหรับสคิล `/implement`, `/check`, `/00-explore` เพื่อส่งเฉพาะข้อมูลที่จำเป็นและควบคุม Token Budget ลดการใช้ Token 60–70%
 
 ---
 
-## ⚡ Phase 2: Feature Expansion & Integrations
+## 🔄 Phase 5: State Drift Detection & Self-Healing
 
-- [ ] **4. Automated Quality Gates & CI Pipeline** `[Size: M]`
-  - *Dependencies*: Phase 1
-  - *Scope*: Wire up `nexus-devflow check-gate`, GitHub Actions workflows, and pre-commit hooks.
-- [ ] **5. External Services & Tooling** `[Size: L]`
+- [x] **5. Git Diff Drift Reconciler & Self-Healing State Engine** `[Size: M]`
   - *Dependencies*: Feature 4
-  - *Scope*: Integrate third-party APIs, MCP servers, or background worker jobs.
+  - *Scope*: ตรวจจับความไม่สอดคล้องระหว่าง `git diff` กับรายการไฟล์ใน Living Spec พร้อมระบบแจ้งเตือนและ Auto-Reconcile ซิงค์สถานะอัตโนมัติ
 
 ---
 
-## 📦 Phase 3: Hardening & Production Release
+## 🖥️ Phase 6: IDE Native Extension & Visual Studio
 
-- [ ] **6. Security & Performance Audit** `[Size: M]`
-  - *Dependencies*: Phase 2
-  - *Scope*: Run `/audit`, resolve P0/P1 blockers, and optimize asset bundle sizes.
-- [ ] **7. Production Deployment & Cloud Smoke Tests** `[Size: S]`
+- [x] **6. IDE Native Extension & Interactive Webview Studio** `[Size: L]`
+  - *Dependencies*: Feature 5
+  - *Scope*: พัฒนา Extension สำหรับ VS Code / Antigravity Webview เพื่อแสดง Live Kanban Board, ปุ่มกดรันคำสั่ง DevFlow และ Diff Inspector ในตัว IDE
+
+
+---
+
+## 🤖 Phase 7: Multi-Agent Swarm Orchestration & Code Graph RAG
+
+- [x] **7. Multi-Agent Swarm Orchestrator & Semantic Code Graph RAG** `[Size: L]`
   - *Dependencies*: Feature 6
-  - *Scope*: Configure Render / Vercel deployment and run end-to-end smoke tests.
+  - *Scope*: พัฒนาระบบกระจายงานแบบคู่ขนานให้ Subagents (Coder, QA Verifier, Security Auditor) และระบบ Indexing Local Codebase Graph เพื่อดึงบริบทที่แม่นยำสูงสุด
+
+---
+
+## 🔄 Phase 8: Upstream Synchronization & Multi-Adapter Expansion
+
+- [x] **8. Sync Upstream v0.13.0 (Adopt Visibility, OpenCode & Multi-Adapter Selection)** `[Size: M]`
+  - *Dependencies*: None
+  - *Scope*: ผสานความสามารถ Adopt Workflow Visibility (Commit vs Local-only), การรองรับ OpenCode และ Multi-Adapter Checkbox Prompt ใน CLI พร้อมอัปเดต Doctor checks และซิงก์ Baseline SHA เป็น v0.13.0 (`0b65166`)
+

package/template/devflow/history/features/README.md

@@ -0,0 +1,5 @@
+# Completed Features Archive
+
+Finished feature specs, enhancements, architecture migrations, and tooling runs are archived here as `xxx-name.md` (for Fast-Track single living specs) or `xxx-name/` (for Deep-Track stage runs).
+
+Together they form the project's permanent feature build history, recording what was built, why, and how it was verified.

package/template/devflow/history/fixes/README.md

@@ -0,0 +1,5 @@
+# Completed Fixes Archive
+
+Finished bug fixes, hotfixes, regressions, security patches, and performance optimizations are archived here as `xxx-name.md` (for Fast-Track single living specs) or `xxx-name/` (for Deep-Track stage runs).
+
+Together they form the project's permanent bug resolution history, recording root causes, TDD test decisions, and empirical proof.

package/template/devflow/history/rollbacks/README.md

@@ -0,0 +1,5 @@
+# Completed Rollbacks Archive
+
+Safely reversed features and rollback execution records are archived here as `YYYY-MM-DD-xxx-name.md` (or `YYYY-MM-DD-xxx-name/`).
+
+Together they form the project's safe feature reversal ledger, documenting target commits, dependency review, and post-reversal verification.

package/template/devflow/project-plan.md

@@ -1,27 +1,79 @@
 # 🗺️ Project Plan (User-Owned Architectural Vision)
 
 > **Document Type**: Project Plan (User-Owned)  
-> **Purpose**: Single source of truth for high-level product vision, architecture, constraints, and non-goals. Distilled into `devflow/context/project-overview.md` by `/overview`.
+> **Purpose**: แหล่งความจริงหลัก (Single Source of Truth) สำหรับวิสัยทัศน์ผลิตภัณฑ์ สถาปัตยกรรมระบบ มาตรการความปลอดภัย กฎเกณฑ์การส่งมอบ และแผนงานพัฒนาต่อยอด ถูกนำไปประมวลผลเป็น `devflow/context/project-overview.md` ผ่านคำสั่ง `/overview`
 
 ---
 
 ## 1. Product Vision & Problem Statement
-- **What this is**: A description of the project and the core problem it solves.
-- **Target Audience**: Who will use this system and why.
-- **Core Value Proposition**: Key differentiator and value delivered.
+
+- **What this is**: **Nexus-DevFlow** คือระบบ **Agentic Workflow Layer & Pair Programming Operating System** ที่เชื่อมต่อระหว่างนักพัฒนา (Human Developer) และปัญญาประดิษฐ์ (AI Coding Agents เช่น Google Antigravity, Claude Code, OpenAI Codex, Cursor, GitHub Copilot) เพื่อยกระดับการพัฒนาซอฟต์แวร์สู่มาตรฐานระดับวิศวกรรมสากล
+- **Target Audience**:
+  - นักพัฒนาเดี่ยว (Solo Developers) ที่ต้องการความเร็วสูงแต่ยังคงมีระเบียบวินัยและหลักฐานการทดสอบที่ชัดเจน
+  - ทีมพัฒนาและองค์กร (Engineering Teams & Tech Leads) ที่ต้องการมาตรฐานการส่งมอบโค้ดที่ตรวจสอบได้ (Traceable), ตรวจสอบย้อนหลังได้ (Auditable) และไม่มีปัญหาโค้ดตีกัน (Zero Merge Conflict)
+- **Core Value Proposition**:
+  - **The 3-Pillars Model**: จัดระเบียบบริบทเป็น 3 เสาหลักอย่างชัดเจน — *🔮 Future (Inbox `ideas.md`)* ➔ *⚡ Present (Living Context `devflow/context/`)* ➔ *📦 Past (Categorized History `devflow/history/`)*
+  - **Dual-Track Delivery Model**: เลือกความเร็วให้เหมาะกับขนาดงาน — *Fast-Track (4 ขั้นตอน)* สำหรับ 85% ของงานประจำวัน และ *Deep-Track (8 ขั้นตอน)* สำหรับ 15% ของงานสถาปัตยกรรม
+  - **Hard Quality Gates**: ระบบปฏิเสธการปล่อยงานเมื่อไม่ผ่านเกณฑ์การทดสอบ หรือมีข้อบกพร่องความปลอดภัยค้างอยู่
+
+---
 
 ## 2. Architecture & Tech Stack
-- **Runtime / Framework**: Primary language, framework, and runtime (e.g. Node.js, Next.js, FastAPI, Go).
-- **Storage / Database**: Data persistence layers, caching, and state storage.
-- **Key Modules / Services**: High-level module breakdown and directory responsibilities.
-- **Integration Points**: External APIs, MCP servers, or third-party services.
-
-## 3. Constraints & Non-Goals
-- **Non-Goals**: What this project explicitly will NOT do in the current scope.
-- **Technical Constraints**: Performance targets, security standards, offline requirements, or backward compatibility rules.
-- **Delivery Rules**: Dual-Track workflow discipline (Fast-Track for standard features, Deep-Track for epics).
-
-## 4. Key Milestones & Roadmap
-- **Milestone 1**: Core Foundation & MVP
-- **Milestone 2**: Feature Expansion & Integrations
-- **Milestone 3**: Enterprise Hardening & Release Polish
+
+- **Runtime & Language**: Node.js (>=18.17.0), TypeScript (Strict ESM-First `"type": "module"`)
+- **Persistence & State**:
+  - Markdown-First Architecture พร้อม Type-Safe Schema Validation (Zod)
+  - Git-Backed History Archive (`devflow/history/features/`, `devflow/history/fixes/`, `devflow/history/rollbacks/`, `HISTORY.md`)
+  - Local State Management & Backup Snapshots (`.nexus/`)
+- **Key Modules & Subsystems**:
+  - **Distribution CLI & Package**: `packages/create-nexus-devflow` (Scaffolding, Doctor, Status, Update, Gatekeeper)
+  - **Interactive Web Dashboard**: Embedded Local Webview Dashboard (`/dashboard`) พร้อม Dual-Track Visualizer และ Real-time Snapshot
+  - **Gatekeeper Engine**: ระบบตรวจสอบคุณภาพโค้ดและ Finding Blockers (`nexus-devflow check-gate`)
+  - **Model Context Protocol (MCP) Server**: JSON-RPC Hub สำหรับเชื่อมต่อ AI Agents แบบ Type-Safe
+  - **Dynamic Context Slicer**: ระบบจัดสรรและตัดตอนบริบทแบบ Just-In-Time (JIT) เพื่อประหยัด Token
+- **Integration Points & AI Adapters**:
+  - Google Antigravity & OpenAI Codex (`.agents/skills/`)
+  - Claude Code (`.claude/skills/`)
+  - Model Context Protocol (`@modelcontextprotocol/sdk`)
+
+---
+
+## 3. Constraints, Defensive Guardrails & Non-Goals
+
+- **5 เสาหลักมาตรการป้องกันความเสี่ยง (Hard Defensive Guardrails)**:
+  1. *Hard Quality Gates & Git Hooks*: บล็อกการ Commit ทันทีหาก Spec ไม่ครบ หรือมี `P0/P1 Finding` ค้างอยู่ในสถานะ `open`
+  2. *Model Context Protocol (MCP)*: บังคับใช้ Typed Function Calling เพื่อป้องกันความเสียหายจากการ Parse ข้อความ Markdown
+  3. *Branch-Scoped Context Isolation*: แยกโฟลเดอร์ State ตาม Git Branch ใน `.nexus/branches/<name>/` ป้องกัน Merge Conflict เมื่อทำงานเป็นทีม
+  4. *Dynamic Context Slicing*: ส่งเฉพาะบริบทที่จำเป็นต่อ Stage นั้นๆ เข้าสู่ AI ลด Token Bloat 60–70%
+  5. *State Drift Reconciler*: ตรวจจับความไม่สอดคล้องระหว่าง Git Diffs กับ Spec Tasks และเสนอทางเลือก Auto-Sync ทันที
+- **Non-Goals**:
+  - ไม่ใช่ Proprietary IDE หรือ Closed Cloud SaaS — DevFlow เป็น Transparent Git-Native Layer ที่ฝังตัวในโปรเจกต์ใดก็ได้
+  - ไม่ใช่ Auto-GPT ที่ปล่อยให้ AI เขียนโค้ดตามใจโดยไม่มี Human-in-the-Loop Review Gates
+- **Delivery Rules**:
+  - เคร่งครัดใน TDD Mandate: โค้ดที่เพิ่มหรือแก้ไขต้องมี Automated Unit Tests เสมอ
+  - ห้ามข้ามขั้นตอน Verification: ต้องมีหลักฐานเชิงประจักษ์ (Empirical Proof) ใน `/check` ก่อนอนุมัติ `/complete`
+
+---
+
+## 4. Key Milestones & Evolution Roadmap
+
+- **🚀 Milestone 1: Hard Quality Gates & Pre-commit Enforcement**
+  - ติดตั้ง `check-gate` เข้าสู่ Git Pre-commit Hooks และ CI Pipeline
+  - บล็อกการส่งมอบเมื่อมี Unchecked Tasks หรือ P0/P1 Blockers
+- **⚡ Milestone 2: DevFlow MCP Server Hub & Type-Safe Schemas (`IDEA-004`)**
+  - พัฒนา MCP Server ผ่าน JSON-RPC สำหรับ Antigravity, Claude, Codex, Cursor
+  - ใช้ Zod Schema ควบคุมการอัปเดต Task, Findings, และ Stage State 100% Type-safe
+- **🌿 Milestone 3: Branch-Scoped Context Isolation & Dynamic Router**
+  - รองรับการเก็บ State แยกตาม Git Branch ป้องกัน Merge Conflict ในทีม
+  - ระบบ Auto-Cleanup ลบ Branch State หลังจบ `/complete` และย้ายเข้าสู่ History
+- **✂️ Milestone 4: JIT Context Slicing & Token Budgeting Engine**
+  - ปรับระบบส่ง Context ให้ AI ตาม Stage แบบไดนามิก ประหยัด Token 70%
+  - ควบคุมขนาด `project-overview.md` ไม่ให้เกินงบประมาณ Token
+- **🔄 Milestone 5: Git Diff Reconciler & Self-Healing State**
+  - ตรวจจับความคลาดเคลื่อนระหว่างการแก้โค้ดของมนุษย์กับ Living Spec
+  - ระบบ Smart Reconciliation แจ้งเตือนและซิงค์สเปกอัตโนมัติ
+- **🖥️ Milestone 6: IDE Native Extension & Visual Kanban Studio**
+  - ส่วนขยายบน VS Code / Antigravity Webview สำหรับควบคุม DevFlow ครบวงจรใน IDE
+- **🤖 Milestone 7: Multi-Agent Swarm Orchestration & Semantic Code RAG**
+  - ระบบประสานงาน Subagents หลายตัว (Coder, QA, Auditor) ทำงานขนานกัน
+  - Local Dependency Graph & Semantic Code Search เพื่อดึงบริบทที่แม่นยำสูงสุด