Local SE Agent Masterclass

{
  "id": "unit_test_synthesizer",
  "name": "Enterprise Unit Test Synthesizer",
  "description": "สร้าง Unit Test คุณภาพระดับ Enterprise ตามมาตรฐาน AAA Pattern และครอบคลุมทั้ง Happy, Boundary และ Exception Paths",
  "triggers": ["test", "unit test", "junit", "pytest", "test case", "mock"],
  "version": "1.0.0"
}

# Role: Senior QA Automation & Test Architect
You are operating in specialized 'Unit Test Synthesizer' mode.

## Mandatory Rules:
1. **Strict AAA Structure**: Every test MUST be explicitly divided into:
   - `// Arrange`: Setup mocks, inputs, and expected values.
   - `// Act`: Execute the single target method under test.
   - `// Assert`: Verify return values, exceptions, and verify mock interactions.
2. **Mocking Isolation**: Never call real external databases or network services. Mock all repositories and clients.
3. **Naming Standard**: Test method names must strictly follow `should_[ExpectedBehavior]_When_[StateUnderTest]()`.
4. **Three Critical Paths**: Every tested function must have:
   - Happy Path: Valid inputs, expected successful outcome.
   - Boundary Path: Empty collections, null checks, edge values (0, MAX_INT).
   - Exception Path: Verify expected domain exceptions with `assertThrows`.

{
  "id": "git_release_manager",
  "name": "Git Release & Changelog Manager",
  "description": "วิเคราะห์ Git Commits สร้าง Release Notes ตามมาตรฐาน Keep a Changelog และจัดการ Semantic Versioning",
  "triggers": ["git release", "changelog", "release notes", "version bump", "tag"],
  "version": "1.0.0"
}

import json
from pathlib import Path

SKILLS_DIR = Path("./skills").resolve()

def estimate_tokens(text: str) -> int:
    """ประมาณการจำนวนโทเค็นอย่างปลอดภัย (Heuristic: ~3.5 ตัวอักษรต่อ 1 token)"""
    return int(len(text) / 3.5)

def simulate_monolithic_prompt(skills_root: Path) -> tuple[str, int]:
    """จำลองวิธีแบบเดิม (Monolithic): ยัดทุกคู่มือตัวเต็มของทุกสกิลลง Prompt ก้อนเดียว"""
    monolithic_parts = [
        "You are an Autonomous Software Engineering Agent.\n",
        "--- GENERAL REACT RULES (500 words) ---\n" + ("Rules..." * 50) + "\n"
    ]
    for skill_folder in skills_root.iterdir():
        if skill_folder.is_dir() and (skill_folder / "instructions.md").exists():
            inst = (skill_folder / "instructions.md").read_text(encoding="utf-8")
            monolithic_parts.append(f"\n--- SKILL MANUAL: {skill_folder.name} ---\n{inst}\n")
            
    full_prompt = "".join(monolithic_parts)
    return full_prompt, estimate_tokens(full_prompt)

def simulate_two_stage_skills(skills_root: Path, active_skill_id: str) -> tuple[str, str, int]:
    """จำลองวิธีใหม่ (Two-Stage Progressive Disclosure): ส่งสารบัญย่อ + ขยายเฉพาะสกิลที่ Active"""
    # Stage 1: สร้าง Catalog สารบัญย่อ
    catalog_items = []
    active_instructions = ""

    for skill_folder in skills_root.iterdir():
        if skill_folder.is_dir() and (skill_folder / "metadata.json").exists():
            meta = json.loads((skill_folder / "metadata.json").read_text(encoding="utf-8"))
            catalog_items.append(f"- [{meta['id']}]: {meta['description']}")
            if meta['id'] == active_skill_id and (skill_folder / "instructions.md").exists():
                active_instructions = (skill_folder / "instructions.md").read_text(encoding="utf-8")

    stage_1_prompt = (
        "You are an Autonomous Software Engineering Microkernel Agent.\n\n"
        "AVAILABLE SKILLS CATALOG:\n" + "\n".join(catalog_items) + "\n\n"
        "When a specialized task arrives, the corresponding manual will be dynamically expanded."
    )
    
    total_active_context = stage_1_prompt + "\n\nACTIVE SKILL EXPANSION:\n" + active_instructions
    return stage_1_prompt, total_active_context, estimate_tokens(total_active_context)

def main():
    print("=" * 70)
    print("🚀 Day 15 Lab: Architecture Blueprint — Dynamic Skills vs. Monolithic Prompts")
    print("=" * 70)

    # 1. ทดสอบจำลอง Monolithic
    _, mono_tokens = simulate_monolithic_prompt(SKILLS_DIR)
    print(f"\n❌ [Monolithic Architecture] ขนาด Context ตั้งต้น: ~{mono_tokens} Tokens")
    print("   👉 ข้อเสีย: ยัดทุกคู่มือพร้อมกันทั้งหมด ยิ่งเพิ่มสกิล Token ยิ่งระเบิด โมเดลสับสน!")

    # 2. ทดสอบจำลอง Two-Stage Progressive Disclosure
    stage1_prompt, active_context, dynamic_tokens = simulate_two_stage_skills(SKILLS_DIR, "unit_test_synthesizer")
    print(f"\n✅ [Dynamic Skills Architecture] ข้อมูล Stage 1 Catalog:")
    print("-" * 50)
    print(stage1_prompt)
    print("-" * 50)
    print(f"🎯 ขนาด Context เมื่อขยายเฉพาะสกิล 'unit_test_synthesizer': ~{dynamic_tokens} Tokens")

    saved_percent = ((mono_tokens - dynamic_tokens) / mono_tokens) * 100
    print(f"\n📊 ผลลัพธ์: ประหยัดพื้นที่ Token เริ่มต้นได้กว่า {saved_percent:.1f}%!")
    print("🏆 พิสูจน์แล้วว่าระบบ Microkernel ช่วยให้ Mac M3 ประมวลผลได้เร็วเฉียบคม ไร้ Prompt Dilution 100%!")

if __name__ == "__main__":
    main()