migrating-agent-systems

# Migrating Agent Systems

## Overview

**Migrating agent systems IS orchestrating component creation through skills.**

Don't create files directly—invoke the appropriate skill for each component type.

**Core principle:** Skills encode best practices. Direct file creation bypasses them.

**Violating the letter of the rules is violating the spirit of the rules.**

## Task Initialization (MANDATORY)

Before ANY action, create task list using TaskCreate:

```
TaskCreate for EACH task below:
- Subject: "[migrating-agent-systems] Task N: <action>"
- ActiveForm: "<doing action>"
```

**Tasks:**
1. Project discovery
2. Architecture planning
3. Create constitution (CLAUDE.md)
4. Create rules
5. Create hooks
6. Create skills (if needed)
7. Validate system

Announce: "Created 7 tasks. Starting execution..."

**Execution rules:**
1. `TaskUpdate status="in_progress"` BEFORE starting each task
2. `TaskUpdate status="completed"` ONLY after verification passes
3. If task fails → stay in_progress, diagnose, retry
4. NEVER skip to next task until current is completed
5. At end, `TaskList` to confirm all completed

## Task 1: Project Discovery

**Goal:** Understand the project before configuring agent system.

**Invoke `project-discovery` skill** (if available) or manually:

**Discover:**
- Tech stack (language, framework, build tools)
- Project structure (key directories)
- Existing workflows (CI/CD, testing, linting)
- Conventions (naming, architecture patterns)

**Output:** Discovery report with key findings.

**Verification:** Have documented tech stack and structure.

## Task 2: Architecture Planning

**Goal:** Design the agent system components.

**Invoke `agent-architect` skill** (if available) or manually plan:

### Component Planning

| Component | Purpose | Needed? |
|-----------|---------|---------|
| CLAUDE.md | Constitution with laws | **Always** |
| Rules | Path-scoped conventions | If conventions exist |
| Hooks | Quality gates | If linting/formatting needed |
| Skil