custom-agent-design

# Custom Agent Design Skill

Design domain-specific agents using Claude Agent SDK patterns.

## Purpose

Guide the design of custom agents that solve domain-specific problems with full SDK control over Context, Model, Prompt, and Tools.

## When to Use

- Building a new custom agent
- Converting generic workflow to specialized agent
- Designing domain-specific automation
- Creating repeatable agent patterns

## Prerequisites

- Clear understanding of the domain problem
- Access to Claude Agent SDK documentation
- Understanding of when custom agents are appropriate (see @agent-evolution-path.md)

## Design Process

### Step 1: Define Agent Purpose

Answer these questions:

- What specific problem does this agent solve?
- What domain expertise does it need?
- What is the single purpose? (One agent, one purpose)
- Who are the stakeholders?

**Output**: Purpose statement (2-3 sentences)

### Step 2: Select Model

Choose based on task complexity:

| Task Type | Model | Why |
| --- | --- | --- |
| Simple transformations | Haiku | Fast, cheap |
| Balanced tasks | Sonnet | Good trade-off |
| Complex reasoning | Opus | Highest quality |

**Decision factors**:

- Speed requirements
- Quality requirements
- Cost constraints
- Task complexity

### Step 3: Design System Prompt

Choose architecture:

**Override** (`system_prompt=...`)

- When: Building a new product
- Result: NOT Claude Code anymore
- Full control over behavior

**Append** (`append_system_prompt=...`)

- When: Extending Claude Code
- Result: Enhanced Claude Code
- Adds capabilities

**System Prompt Template**:

```markdown
# [Agent Name]

## Purpose
[Identity and role definition - 2-3 sentences]

## Instructions
[Core behaviors - bullet list]
- Behavior 1
- Behavior 2

## Constraints
[What the agent must NOT do]

## Examples (if needed)
[Input/Output pairs]
```

### Step 4: Configure Tool Access

**Questions to answer**:

- What tools does this agent need?
- What tools should be blocked?
- Are custom tools requir