thachvd-kit 1.0.0

package/kit/PROMPT_RECIPE.md

@@ -0,0 +1,52 @@
+# 📜 PROMPT RECIPES - Coding Agent Strategy
+
+Use these templates to enforce modular design, clean architecture, and contextual awareness in any project.
+
+---
+
+## 🏗️ 1. Project Initiation (Architecture First)
+Use this when starting a new project. Force the Agent to design the structure before writing any code.
+
+> "I want to initialize a new project: [Project Name]. 
+> **Stack:** [e.g., Next.js, FastAPI, Flutter, etc.]
+> 
+> **Architectural Requirements:**
+> 1. Follow a **Modular & Scalable** approach. Keep components atomic and logic decoupled.
+> 2. **State Management:** Use [e.g., Zustand, Redux, Provier, or none] with a modular pattern.
+> 3. **Separation of Concerns:** Keep business logic outside of UI files (use services/use-cases/lib).
+> 4. **Resource Management:** Offload heavy computations or system-level tasks to [e.g., Rust, Go, or Backend].
+> 5. **Constraints:** Enforce a maximum of **250 lines per file** to ensure atomicity.
+> 
+> Please propose a detailed **File Tree** and explain the responsibility of each directory before implementation."
+
+---
+
+## ⚡ 2. Complex Feature Implementation
+Use this for hard tasks that require deep thinking.
+
+> "I need to implement the following feature: [Feature Description].
+> 
+> **Execution Workflow:**
+> 1. Analyze the project baseline from `PROJECT_CONTEXT.md`.
+> 2. Propose a **Data Flow Diagram** or a sequence of operations.
+> 3. Break down the task into small, manageable files: State/Store, Logic/Service, and UI/Component.
+> 4. Ensure compliance with the project's **Clean Code** standards."
+
+---
+
+## 🐞 3. Systemic Debugging
+Use this when facing logic errors or unexpected behavior.
+
+> "The application is encountering the following issue: [Error Description].
+> 
+> **Instructions:**
+> 1. Identify the **Root Cause** by inspecting logs or using diagnostic tools.
+> 2. Verify if this issue impacts the system assumptions defined in `PROJECT_CONTEXT.md`.
+> 3. Fix the issue using a sustainable approach. If an architectural refactor is required for a permanent fix, please propose it first."
+
+---
+
+## 🔄 4. Context Synchronization
+Use this at the end of a session or after significant structural changes.
+
+> "I have completed [Task/Feature]. Please update the **`PROJECT_CONTEXT.md`** file to reflect the latest state of the project, including new files created, technical decisions made, and current progress status."

package/kit/README.md

@@ -0,0 +1,32 @@
+# 🛡️ Antigravity Project Kit
+
+This toolkit helps you build high-quality, modular, and performance-optimized applications by leveraging the full power of AI Agents with a Context-First approach.
+
+---
+
+## 🚀 How to Use
+
+### 1. Initialization (Clone & Use)
+- Clone this repository.
+- Update project metadata (e.g., `package.json`).
+- Rename and fill out `PROJECT_CONTEXT.template.md` as `PROJECT_CONTEXT.md` in the root directory.
+
+### 2. Working with AI Agents
+- Always use the templates provided in `kit/PROMPT_RECIPE.md`.
+- Ensure the Agent reads and updates `PROJECT_CONTEXT.md`.
+- Enforce file splitting rules (e.g., Max 250 lines) to maintain a clean codebase.
+
+### 3. Proposed Structure
+- `.agent/rules/GEMINI.md`: Core behavioral rules and protocols.
+- `.agent/kit/`: Manuals and templates for project management.
+- `PROJECT_CONTEXT.md`: (Manual creation required) The active map of the current project.
+
+## 💎 Golden Rules
+1. **Context is King:** Never let the Agent work without a clear Ground Truth.
+2. **Architecture First:** Design the structure before writing any implementation code.
+3. **Keep it Small:** Small files are easier for AI to understand, modify, and keep in context.
+4. **Self-Documenting:** The Agent is responsible for maintaining the project documentation.
+
+---
+
+*Powered by Antigravity Framework.*

package/mcp_config.json

@@ -0,0 +1,24 @@
+{
+    "mcpServers": {
+        "context7": {
+            "command": "npx",
+            "args": [
+                "-y",
+                "@upstash/context7-mcp",
+                "--api-key",
+                "YOUR_API_KEY"
+            ]
+        },
+        "shadcn": {
+            "command": "npx",
+            "args": [
+                "shadcn@latest",
+                "mcp"
+            ]
+        }
+        //other mcp servers
+    }
+}
+
+// setup mcp server in  ~/.gemini\antigravity\mcp_config.json
+

package/package.json

@@ -0,0 +1,25 @@
+{
+    "name": "thachvd-kit",
+    "version": "1.0.0",
+    "description": "Universal Agent Kit for Antigravity, Claude Code, and OpenCode",
+    "main": "index.js",
+    "bin": {
+        "thachvd-kit": "./bin/cli.js"
+    },
+    "scripts": {
+        "test": "echo \"Error: no test specified\" && exit 1"
+    },
+    "keywords": [
+        "ai",
+        "agent",
+        "claude",
+        "antigravity",
+        "codex"
+    ],
+    "author": "thachvd",
+    "license": "MIT",
+    "dependencies": {
+        "picocolors": "^1.0.0",
+        "prompts": "^2.4.2"
+    }
+}

package/rules/GEMINI.md

@@ -0,0 +1,204 @@
+---
+trigger: always_on
+---
+
+# GEMINI.md — Universal Agent Kit
+
+> Cross-platform agent protocol. Works with: Antigravity · Claude Code · Codex · OpenCode
+
+---
+
+## MANDATORY FIRST ACTION
+
+**At session start — before anything else:**
+
+```
+1. Read PROJECT_CONTEXT.md (if exists) → load stack, agents, constraints
+2. If missing → run @[skills/project-onboarding] to create it
+3. Confirm: "Context loaded: [Project Name] | Stack: [Stack] | Ready."
+```
+
+> All routing decisions below depend on PROJECT_CONTEXT.md.
+> Without it, ask user to run: `python scripts/init.py`
+
+---
+
+## STEP 1 — CLASSIFY REQUEST
+
+| Type | Triggers | Action |
+|------|----------|--------|
+| **QUESTION** | "what", "how", "explain", "why" | Direct answer |
+| **SURVEY** | "analyze", "list", "overview", "check" | Analysis only, no code |
+| **SIMPLE FIX** | "fix", "add", "change" (single file) | Inline edit |
+| **BUILD** | "build", "create", "implement", "refactor" | Agent + `{task-slug}.md` |
+| **DESIGN/UI** | "design", "UI", "page", "dashboard" | Agent + `{task-slug}.md` |
+
+---
+
+## STEP 2 — AUTO-SELECT AGENT
+
+**Protocol:**
+1. Read `stack.primary_language` and `stack.frameworks` from PROJECT_CONTEXT.md
+2. Match to agent table below
+3. Announce before responding:
+
+```
+🤖 Applying knowledge of `@[agent-name]`...
+```
+
+### Agent Routing Table
+
+| Domain | Agent | Skills to Load |
+|--------|-------|---------------|
+| Web frontend (any framework) | `frontend-specialist` | `frontend-design` + framework skill from PROJECT_CONTEXT |
+| Backend API (any language) | `backend-specialist` | language skill + `api-design` |
+| Database / Schema | `database-architect` | `database-design` |
+| Mobile / Desktop (Tauri, RN, Flutter) | `mobile-developer` | `mobile-design` |
+| DevOps / Cloud / CI | `devops-engineer` | `docker-patterns`, `deployment-procedures` |
+| Security audit | `security-auditor` | `vulnerability-scanner` |
+| Debug / RCA | `debugger` | `systematic-debugging` |
+| Planning / Discovery | `project-planner` | `brainstorming`, `plan-writing` |
+| Multi-domain | `orchestrator` | `dispatching-parallel-agents` |
+
+**Framework-specific skill mapping** (reference PROJECT_CONTEXT.stack):
+
+| Framework/Language | Load Skill |
+|-------------------|-----------|
+| Vue | `frontend-design` (+ Vue-specific if available) |
+| React / Next.js | `nextjs-react-expert` |
+| Laravel | `laravel-patterns` + `laravel-security` |
+| Node.js | `nodejs-best-practices` |
+| Python | `python-patterns` |
+| Go | `golang-patterns` |
+| Rust / Tauri | `rust-pro` |
+| Docker | `docker-patterns` |
+
+> If stack not in mapping → use agent's general skill set + `clean-code`.
+
+**MANDATORY checklist before any code:**
+- [ ] Agent identified from PROJECT_CONTEXT.md stack?
+- [ ] Agent `.md` file read/recalled?
+- [ ] Announcement `🤖 Applying knowledge of @[agent]...` written?
+
+> 🔴 Code without announcement = **PROTOCOL VIOLATION**
+
+---
+
+## STEP 3 — SOCRATIC GATE
+
+**Trigger: BUILD or DESIGN/UI requests only**
+
+| Scenario | Action |
+|----------|--------|
+| New feature / greenfield | Ask **3 questions**: Purpose · Users · Must-have vs Nice-to-have |
+| Code edit / bug fix | Confirm scope + 1 impact question |
+| Vague request | Ask Purpose + Constraints |
+| User gave full spec | **Proceed immediately** |
+| User says "proceed" / "just do it" | **DO IT — no more questions** |
+
+**Rules:**
+- Max 3 questions at once
+- Never ask edge case questions after user already answered
+- Reference `@[skills/brainstorming]` for question format
+
+---
+
+## UNIVERSAL RULES (Always Active)
+
+### Language
+- Respond in user's language (auto-detect from prompt)
+- Code comments/variables always in English
+
+### Code Quality
+- `@[skills/clean-code]` applies to ALL code, ALL languages
+- Tests mandatory: Unit > Integration > E2E
+- `@[skills/verification-before-completion]` before claiming "done"
+
+### File Changes
+- Identify dependent files before modifying
+- Update ALL affected files together
+
+---
+
+## EXECUTION WORKFLOW
+
+**BUILD tasks — 7-step flow:**
+
+```
+1. brainstorming           → Clarify (Socratic Gate)
+2. plan-writing            → Create {task-slug}.md
+3. using-git-worktrees     → Isolated workspace
+4. subagent-driven-development → Implement + 2-stage review
+5. requesting-code-review  → After each major task
+6. verification-before-completion → Before claiming done
+7. finishing-a-development-branch → Merge / PR / keep
+```
+
+**SIMPLE FIX:** Skip to step 4, no worktree needed.
+
+---
+
+## MODE BEHAVIOR
+
+| Mode | Behavior |
+|------|----------|
+| **plan** | Analysis → Planning → Architecture → NO CODE until confirmed |
+| **ask** | Questions and analysis only |
+| **edit** | Execute. Multi-file → create `{task-slug}.md`. Single-file → proceed. |
+
+---
+
+## FINAL CHECK
+
+**Trigger:** "final check", "deploy check", "kiểm tra cuối", "done?"
+
+```bash
+python scripts/checklist.py .
+```
+
+Priority: Security → Lint → Tests → UX → SEO
+
+---
+
+## PATHS
+
+```
+agents/     → agents/{agent}.md
+skills/     → skills/{skill}/SKILL.md
+workflows/  → workflows/{command}.md
+scripts/    → scripts/init.py, checklist.py, verify_all.py
+```
+
+---
+
+## QUICK SKILL REFERENCE
+
+| Need | Skill |
+|------|-------|
+| Clean code (any lang) | `clean-code` |
+| Plan a task | `plan-writing` → `executing-plans` |
+| Parallel agents | `dispatching-parallel-agents` |
+| Code review | `requesting-code-review` / `receiving-code-review` |
+| Git isolation | `using-git-worktrees` |
+| Agent-based dev | `subagent-driven-development` |
+| Verify done | `verification-before-completion` |
+| Finish branch | `finishing-a-development-branch` |
+| Write new skill | `writing-skills` |
+| API design | `api-design` |
+| Docker | `docker-patterns` |
+| Laravel | `laravel-patterns`, `laravel-security`, `laravel-tdd` |
+| Go | `golang-patterns`, `golang-testing` |
+| Debugging | `systematic-debugging` |
+| Security | `vulnerability-scanner` |
+
+---
+
+## CROSS-PLATFORM NOTES
+
+| Platform | Rules File | Skills Dir | Workflows |
+|----------|-----------|-----------|-----------|
+| **Antigravity** | `rules/GEMINI.md` (this file) | `skills/` | `workflows/` |
+| **Claude Code** | `CLAUDE.md` at project root | `.claude/skills/` | `.claude/commands/` |
+| **Codex / OpenCode** | `CLAUDE.md` at project root | `.claude/skills/` | — |
+
+> Run `python scripts/init.py --platform all` to generate platform-specific config files automatically.

package/scripts/auto_preview.py

@@ -0,0 +1,148 @@
+#!/usr/bin/env python3
+"""
+Auto Preview - Antigravity Kit
+==============================
+Manages (start/stop/status) the local development server for previewing the application.
+
+Usage:
+    python .agent/scripts/auto_preview.py start [port]
+    python .agent/scripts/auto_preview.py stop
+    python .agent/scripts/auto_preview.py status
+"""
+
+import os
+import sys
+import time
+import json
+import signal
+import argparse
+import subprocess
+from pathlib import Path
+
+AGENT_DIR = Path(".agent")
+PID_FILE = AGENT_DIR / "preview.pid"
+LOG_FILE = AGENT_DIR / "preview.log"
+
+def get_project_root():
+    return Path(".").resolve()
+
+def is_running(pid):
+    try:
+        os.kill(pid, 0)
+        return True
+    except OSError:
+        return False
+
+def get_start_command(root):
+    pkg_file = root / "package.json"
+    if not pkg_file.exists():
+        return None
+    
+    with open(pkg_file, 'r') as f:
+        data = json.load(f)
+    
+    scripts = data.get("scripts", {})
+    if "dev" in scripts:
+        return ["npm", "run", "dev"]
+    elif "start" in scripts:
+        return ["npm", "start"]
+    return None
+
+def start_server(port=3000):
+    if PID_FILE.exists():
+        try:
+            pid = int(PID_FILE.read_text().strip())
+            if is_running(pid):
+                print(f"⚠️  Preview already running (PID: {pid})")
+                return
+        except:
+            pass # Invalid PID file
+
+    root = get_project_root()
+    cmd = get_start_command(root)
+    
+    if not cmd:
+        print("❌ No 'dev' or 'start' script found in package.json")
+        sys.exit(1)
+    
+    # Add port env var if needed (simple heuristic)
+    env = os.environ.copy()
+    env["PORT"] = str(port)
+    
+    print(f"🚀 Starting preview on port {port}...")
+    
+    with open(LOG_FILE, "w") as log:
+        process = subprocess.Popen(
+            cmd,
+            cwd=str(root),
+            stdout=log,
+            stderr=log,
+            env=env,
+            shell=True # Required for npm on windows often, or consistent path handling
+        )
+    
+    PID_FILE.write_text(str(process.pid))
+    print(f"✅ Preview started! (PID: {process.pid})")
+    print(f"   Logs: {LOG_FILE}")
+    print(f"   URL: http://localhost:{port}")
+
+def stop_server():
+    if not PID_FILE.exists():
+        print("ℹ️  No preview server found.")
+        return
+
+    try:
+        pid = int(PID_FILE.read_text().strip())
+        if is_running(pid):
+            # Try gentle kill first
+            os.kill(pid, signal.SIGTERM) if sys.platform != 'win32' else subprocess.call(['taskkill', '/F', '/T', '/PID', str(pid)])
+            print(f"🛑 Preview stopped (PID: {pid})")
+        else:
+            print("ℹ️  Process was not running.")
+    except Exception as e:
+        print(f"❌ Error stopping server: {e}")
+    finally:
+        if PID_FILE.exists():
+            PID_FILE.unlink()
+
+def status_server():
+    running = False
+    pid = None
+    url = "Unknown"
+    
+    if PID_FILE.exists():
+        try:
+            pid = int(PID_FILE.read_text().strip())
+            if is_running(pid):
+                running = True
+                # Heuristic for URL, strictly we should save it
+                url = "http://localhost:3000" 
+        except:
+            pass
+            
+    print("\n=== Preview Status ===")
+    if running:
+        print(f"✅ Status: Running")
+        print(f"🔢 PID: {pid}")
+        print(f"🌐 URL: {url} (Likely)")
+        print(f"📝 Logs: {LOG_FILE}")
+    else:
+        print("⚪ Status: Stopped")
+    print("===================\n")
+
+def main():
+    parser = argparse.ArgumentParser()
+    parser.add_argument("action", choices=["start", "stop", "status"])
+    parser.add_argument("port", nargs="?", default="3000")
+    
+    args = parser.parse_args()
+    
+    if args.action == "start":
+        start_server(int(args.port))
+    elif args.action == "stop":
+        stop_server()
+    elif args.action == "status":
+        status_server()
+
+if __name__ == "__main__":
+    main()