Dispatching Parallel Agents

The Pattern

1. Identify Independent Domains

Group tasks by what's broken/needed:

• Domain A: auth flow tests
• Domain B: payment subsystem
• Domain C: notification queue

Each domain is independent — fixing one doesn't affect others.

2. Craft Focused Agent Prompts

Each agent gets:

• Specific scope: One file, one subsystem, one concern
• Clear goal: Exactly what to produce or fix
• Constraints: What NOT to touch
• Expected output: Summary of what was found and changed

3. Dispatch in Parallel

// Conceptual — use your platform's parallel task API
Task("Fix auth-flow.test.ts failures")
Task("Fix payment-processor.test.ts failures")
Task("Fix notification-queue.test.ts failures")
// All three run concurrently

4. Review and Integrate

When agents return:

1. Read each summary
2. Verify fixes don't conflict
3. Run full test suite
4. Integrate all changes

Agent Prompt Template

Fix the failing tests in [path/to/test.ts]:

Failures:
1. "[test name]" — expects [X], gets [Y]
2. "[test name]" — [description]

Context:
- These are [type of] issues
- Relevant source file: [path]
- Do NOT modify: [other paths]

Your task:
1. Read the test file and understand each failure
2. Read the source file(s) to understand current behavior
3. Implement minimal fixes
4. Run tests to verify: [test command]
5. Return summary: what you found and what you changed

Scope Isolation Rules

• Each agent must be given ONLY the files it needs
• Agents must not share write access to the same file
• If agents need a shared utility, create it first, then dispatch
• Parallel agents should commit separately or report diffs for human integration

Integration

Works well with:

• subagent-driven-development — for plan-based task dispatch
• executing-plans — feeds tasks to parallel agents
• verification-before-completion — verify each agent's result independently