Dispatching Parallel Agents

Adapted from obra/superpowers for Cursor IDE.

Overview

When you have multiple unrelated failures or independent tasks, investigating them sequentially wastes time. Each investigation is independent and can happen in parallel.

Core principle: Dispatch one subagent per independent problem domain via Cursor's Task tool. Let them work concurrently.

When to Use

• •2+ independent problems (different test files, different subsystems)
• •Each problem can be understood without context from others
• •No shared state between investigations

Don't use when:

• •Failures are related (fix one might fix others)
• •Agents would interfere (editing same files)

The Pattern

1. Identify Independent Domains

Group failures by what's broken. Each domain should be independent.

2. Create Focused Subagent Tasks

Each subagent gets:

• •Specific scope: One test file or subsystem
• •Clear goal: Make these tests pass
• •Constraints: Don't change other code
• •Expected output: Summary of findings and fixes

3. Dispatch in Parallel via Cursor's Task Tool

Use multiple Task tool calls in a single message to dispatch concurrently:

Task 1: { subagent_type: "generalPurpose", description: "Fix file-A tests", prompt: "..." }
Task 2: { subagent_type: "generalPurpose", description: "Fix file-B tests", prompt: "..." }
Task 3: { subagent_type: "generalPurpose", description: "Fix file-C tests", prompt: "..." }

Important: Maximum 4 concurrent subagents in Cursor.

4. Review and Integrate

• •Read each summary
• •Verify fixes don't conflict
• •Run full test suite
• •Integrate all changes

Common Mistakes

• •Too broad: "Fix all the tests" → agent gets lost
• •No context: Paste error messages and test names
• •No constraints: Agent might refactor everything
• •Vague output: Specify "Return summary of root cause and changes"