subagent-engineering

# Subagent Engineering

Manage the full lifecycle of Claude Code subagents: creation, evaluation,
iteration, and troubleshooting.

<prerequisite>
**Subagent prompts are system prompts.** Before creating or improving
a subagent, invoke `prompt-engineering` to load instruction design techniques.

```
Skill(ai-helpers:prompt-engineering)
```

Skip only for trivial edits (typos, formatting).
</prerequisite>

## What Are You Trying To Do?

| Goal | Read |
|------|------|
| Create a new subagent | [creation.md](references/creation.md) |
| Evaluate subagent quality | [evaluation.md](references/evaluation.md) |
| Improve an existing subagent | [iteration.md](references/iteration.md) |
| Debug a subagent that isn't working | [troubleshooting.md](references/troubleshooting.md) |
| Check frontmatter rules and constraints | [spec.md](references/spec.md) |
| See common patterns and examples | [patterns.md](references/patterns.md) |
| Apply prompt engineering techniques | `prompt-engineering` skill |

## What Good Looks Like

A well-designed subagent:
- Triggers correctly (not too often, not too rarely)
- Completes tasks without scope creep
- Returns concise, actionable summaries
- Uses minimal tools for its purpose

## Built-in Subagents

| Agent | Model | Purpose |
|-------|-------|---------|
| **Explore** | Haiku | Fast, read-only codebase exploration |
| **Plan** | Inherits | Research for plan mode |
| **general-purpose** | Inherits | Complex multi-step tasks |
| **Bash** | Inherits | Command execution in separate context |
| **claude-code-guide** | Haiku | Questions about Claude Code features |

## How Subagents Work

Subagents run in isolated context with custom system prompts and restricted
tools. Claude delegates based on `name` + `description` only — the body
loads after delegation. See [spec.md](references/spec.md) for technical details.

## Core Principles

**1. Description is the trigger**
Claude sees ONLY `name` and `description` when deciding to delegate.
The body l