parallel-agents

name	parallel-agents
description	Dispatch multiple agents to work on independent problems concurrently. Use when facing 3+ independent failures or tasks.

Dispatching Parallel Agents

Assign separate agents to independent problem domains simultaneously for faster resolution.

When to Use

• 3+ test failures across different files/subsystems
• Multiple independent tasks that don't share state
• Investigations that won't interfere with each other
• Failures from unrelated root causes

When NOT to Use

• Failures are interconnected
• Tasks share state or create conflicts
• Agents would modify the same files
• You lack context to properly scope tasks

Implementation Steps

1. Group by Domain

Organize failures/tasks into independent categories:

Group A: Authentication tests (3 failures)
Group B: API endpoint tests (2 failures)
Group C: UI component tests (4 failures)

2. Define Focused Tasks

Each agent receives:

Field	Description
Scope	Specific files/tests to focus on
Goal	Clear success criteria
Constraints	What NOT to change
Output	Expected deliverable

3. Dispatch Concurrently

IMPORTANT: Launch all tasks in a single message (no run_in_background). Multiple Task calls in the same message automatically run in parallel, and Claude waits for all to complete.

# All three tasks run in parallel automatically when in the same message
Task(test-engineer, prompt="Fix auth test failures in src/auth/*.test.ts")
Task(test-engineer, prompt="Fix API test failures in src/api/*.test.ts")
Task(frontend-developer, prompt="Fix UI test failures in src/components/*.test.tsx")
# Claude waits for all to complete, then continues