references/module-4-agents.md
Module 4: Agent System
Built-in Agents
| Agent | Model | Best For | Key Trait |
|-------|-------|----------|-----------|
| explore | Haiku | Fast codebase Q&A | Read-only, <300 words, safe to parallelize |
| task | Haiku | Running commands (tests, builds, lints) | Brief on success, verbose on failure |
| general-purpose | Sonnet | Complex multi-step tasks | Full toolset, separate context window |
| code-review | Sonnet | Analyzing code changes | Never modifies code, high signal-to-noise |
Custom Agents — define your own in Markdown
| Level | Location | Scope |
|-------|----------|-------|
| Personal | ~/.copilot/agents/*.md | All your projects |
| Project | .github/agents/*.md | Everyone on this repo |
| Organization | .github-private/agents/ in org repo | Entire org |
Agent file anatomy
---
name: my-agent
description: What this agent does
tools:
- bash
- edit
- view
---
# Agent Instructions
Your detailed behavior instructions here.
Agent orchestration patterns
- Fan-out exploration — Launch multiple
exploreagents in parallel to answer different questions simultaneously - Pipeline —
explore→ understand →general-purpose→ implement →code-review→ verify - Specialist handoff — Identify task →
/agentto pick specialist → review with/fleetor/tasks
Key insight: The AI automatically delegates to subagents when appropriate.