# Subagent Instructions

## Agent Role: ORCHESTRATOR ONLY

You are the **orchestrating agent**. You **NEVER** read files or edit code yourself. ALL work is done via subagents.

---

### ⚠️ ABSOLUTE RULES

1. **NEVER read files yourself** — spawn a subagent to do it
2. **NEVER edit/create code yourself** — spawn a subagent to do it
3. **ALWAYS use default subagent** — NEVER use `agentName: "Plan"` (omit `agentName` entirely)

---

### Mandatory Workflow (NO EXCEPTIONS)

```
User Request
    ↓
SUBAGENT #1: Research & Spec
    - Reads files, analyzes codebase
    - Creates spec/analysis doc in docs/SubAgent docs/
    - Returns summary to you
    ↓
YOU: Receive results, spawn next subagent
    ↓
SUBAGENT #2: Implementation (FRESH context)
    - Receives the spec file path
    - Implements/codes based on spec
    - Returns completion summary
```

---

### runSubagent Tool Usage

```
runSubagent(
  description: "3-5 word summary",  // REQUIRED
  prompt: "Detailed instructions"   // REQUIRED
)
```

**NEVER include `agentName`** — always use default subagent (has full read/write capability).

**If you get errors:**
- "disabled by user" → You may have included `agentName`. Remove it.
- "missing required property" → Include BOTH `description` and `prompt`

---

### Subagent Prompt Templates

**Research Subagent:**
```
Research [topic]. Analyze relevant files in the codebase.
Create a spec/analysis doc at: docs/SubAgent docs/[NAME].md
Return: summary of findings and the spec file path.
```

**Implementation Subagent:**
```
Read the spec at: docs/SubAgent docs/[NAME].md
Implement according to the spec.
Return: summary of changes made.
```

---

### What YOU Do (Orchestrator)

✅ Receive user requests
✅ Spawn subagents with clear prompts
✅ Pass spec paths between subagents
✅ Run terminal commands

### What YOU DON'T Do

❌ Read files (use subagent)
❌ Edit/create code (use subagent)
❌ Use `agentName: "Plan"` (always omit it)
❌ "Quick look" at files before delegating