dispatch-agents

name	dispatch-agents
description	Orchestrate multiple subagents in parallel for independent tasks. Use when a feature has multiple unrelated parts that can be worked on simultaneously, or when multiple test failures affect different subsystems.
triggers	["/dispatch","parallel agents","dispatch agents","work in parallel"]
allowed-tools	Read, Glob, Grep, Agent, TodoWrite
user-invocable	true

Dispatching Parallel Agents

Delegate independent tasks to multiple subagents working concurrently instead of solving everything sequentially.

When to Use This

• Multiple unrelated failures across different files/subsystems
• A feature with independent parts (e.g., API + UI + tests that don't depend on each other)
• Investigation tasks that can be scoped to separate domains

When NOT to Use This

• Tasks share state or modify the same files
• Failures are related (fixing one might fix the others)
• You need full system understanding before acting
• Exploratory debugging (use /debug instead)

The Process

Step 1: Identify Independent Domains

Group work by what's broken or what needs building. Each domain must be:

• Self-contained — understandable without the other domains
• Non-overlapping — agents won't edit the same files
• Clear in scope — specific goal, not "fix stuff"

Step 2: Create Focused Agent Prompts

Each agent gets a prompt with:

1. Context — what the project is, what stack is used, relevant CLAUDE.md rules
2. Specific scope — exactly which files/modules to work on
3. Clear goal — what "done" looks like
4. Constraints — what NOT to touch, what patterns to follow

Good prompt:

"Fix the authentication tests in tests/auth.test.ts. The signIn function now returns { user, session } instead of just user (changed in lib/auth.ts line 42). Update the test expectations to match the new return type. Run pnpm test tests/auth.test.ts to verify."

Bad prompt:

"Fix the failing tests."

Step 3: Dispatch in Parallel

Use the Agent tool to launch all agents concurrently in a single message:

Agent 1: [Domain A task with full context]
Agent 2: [Domain B task with full context]
Agent 3: [Domain C task with full context]

Step 4: Review and Integrate

Once all agents complete:

1. Check for conflicts — did any agents modify the same files?
2. Run the full test suite — pnpm test to verify everything works together
3. Review each agent's changes — quick sanity check on quality
4. Fix integration issues — if agents' changes don't compose well

Agent Dispatch Checklist

Before dispatching, verify:

• Tasks are truly independent (no shared file modifications)
• Each prompt has enough context to work autonomously
• Each prompt specifies which files to work on
• Each prompt includes the verification command to run
• No agent needs the output of another agent to start

Practical Examples

Multiple Test Failures (different modules)

Agent 1: "Fix tests in tests/auth.test.ts — signIn return type changed from User to { user, session }"
Agent 2: "Fix tests in tests/billing.test.ts — Stripe API mock needs updating for v2 webhook format"
Agent 3: "Fix tests in tests/dashboard.test.ts — new required prop 'organizationId' on DashboardLayout"

Feature with Independent Parts

Agent 1: "Create the database migration and RLS policies for the 'notifications' table"
Agent 2: "Build the NotificationBell component following the design system"
Agent 3: "Write the Server Action for marking notifications as read"

Limits

• 3-5 agents max — more than that and integration becomes the bottleneck
• Don't dispatch implementation agents in parallel on the same feature if they might conflict — use sequential execution instead
• Always run the full test suite after integration — parallel work can create subtle incompatibilities