---

name: openagent-dispatching-parallel-agents description: Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies - with OpenAgent approval gates

Dispatching Parallel Agents - OpenAgent Version

Overview

When you have multiple unrelated failures (different test files, different subsystems, different bugs), investigating them sequentially wastes time. Each investigation is independent and can happen in parallel. With approval gates at each phase.

Core principle: Dispatch one agent per independent problem domain. Let them work concurrently.

OpenAgent integration: This skill integrates approval gates from OpenAgent's safety-first philosophy. You will request approval before identifying domains, dispatching agents, and integrating changes.

When to Use

digraph when_to_use {
    "Multiple failures?" [shape=diamond];
    "Are they independent?" [shape=diamond];
    "Single agent investigates all" [shape=box];
    "One agent per problem domain" [shape=box];
    "Can they work in parallel?" [shape=diamond];
    "Sequential agents" [shape=box];
    "Parallel dispatch" [shape=box];

    "Multiple failures?" -> "Are they independent?" [label="yes"];
    "Are they independent?" -> "Single agent investigates all" [label="no - related"];
    "Are they independent?" -> "Can they work in parallel?" [label="yes"];
    "Can they work in parallel?" -> "Parallel dispatch" [label="yes"];
    "Can they work in parallel?" -> "Sequential agents" [label="no - shared state"];
}

Use when:

• 3+ test files failing with different root causes
• Multiple subsystems broken independently
• 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)
• Need to understand full system state
• Agents would interfere with each other

The Pattern with Approval Gates

1. Identify Independent Domains

⏸️ REQUEST APPROVAL: "May I analyze these failures to identify independent problem domains?"

After approval, group failures by what's broken:

• File A tests: Tool approval flow
• File B tests: Batch completion behavior
• File C tests: Abort functionality

Each domain is independent - fixing tool approval doesn't affect abort tests.

⏸️ REPORT TO USER: "Identified [N] independent domains: [list domains]. These can be worked in parallel. Ready to proceed?"

2. Create Focused Agent Tasks

⏸️ REQUEST APPROVAL: "May I create agent tasks for these [N] domains? Review scope below:"

After approval, each agent gets:

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

3. Dispatch in Parallel

⏸️ REQUEST APPROVAL: "May I dispatch [N] agents in parallel with the following tasks?"

After approval, dispatch concurrently:

// In Claude Code / AI environment
Task("Fix agent-tool-abort.test.ts failures")
Task("Fix batch-completion-behavior.test.ts failures")
Task("Fix tool-approval-race-conditions.test.ts failures")
// All three run concurrently

4. Review and Integrate

⏸️ REPORT AGENT SUMMARIES: "All agents have returned. Summaries:

• Agent 1: [summary]
• Agent 2: [summary]
• Agent 3: [summary]"

After reporting, verify:

• Read each summary
• Verify fixes don't conflict
• ⏸️ REQUEST APPROVAL: "May I integrate all changes and run the full test suite?"

After approval:

• Run full test suite
• Integrate all changes

⏸️ REPORT RESULTS: "Integration complete. Test results: [show output]"

Agent Prompt Structure

Good agent prompts are:

1. Focused - One clear problem domain
2. Self-contained - All context needed to understand the problem
3. Specific about output - What should the agent return?

Fix the 3 failing tests in src/agents/agent-tool-abort.test.ts:

1. "should abort tool with partial output capture" - expects 'interrupted at' in message
2. "should handle mixed completed and aborted tools" - fast tool aborted instead of completed
3. "should properly track pendingToolCount" - expects 3 results but gets 0

These are timing/race condition issues. Your task:

1. Read the test file and understand what each test verifies
2. Identify root cause - timing issues or actual bugs?
3. Fix by:
   - Replacing arbitrary timeouts with event-based waiting
   - Fixing bugs in abort implementation if found
   - Adjusting test expectations if testing changed behavior

Do NOT just increase timeouts - find the real issue.

Return: Summary of what you found and what you fixed.

Common Mistakes

❌ Too broad: "Fix all the tests" - agent gets lost ✅ Specific: "Fix agent-tool-abort.test.ts" - focused scope

❌ No context: "Fix the race condition" - agent doesn't know where ✅ Context: Paste the error messages and test names

❌ No constraints: Agent might refactor everything ✅ Constraints: "Do NOT change production code" or "Fix tests only"

❌ Vague output: "Fix it" - you don't know what changed ✅ Specific: "Return summary of root cause and changes"

❌ Skipping approval gates: Dispatching agents without user awareness ✅ Request approval: Before identifying domains, creating tasks, dispatching agents

When NOT to Use

Related failures: Fixing one might fix others - investigate together first Need full context: Understanding requires seeing entire system Exploratory debugging: You don't know what's broken yet Shared state: Agents would interfere (editing same files, using same resources)

Real Example with Approval Gates

Scenario: 6 test failures across 3 files after major refactoring

Failures:

• agent-tool-abort.test.ts: 3 failures (timing issues)
• batch-completion-behavior.test.ts: 2 failures (tools not executing)
• tool-approval-race-conditions.test.ts: 1 failure (execution count = 0)

⏸️ REQUEST APPROVAL: "May I analyze these 6 failures to identify independent domains?"

After approval - Analysis: Independent domains - abort logic separate from batch completion separate from race conditions

⏸️ REPORT: "Identified 3 independent domains. Ready to create agent tasks?"

After approval - Dispatch:

Agent 1 → Fix agent-tool-abort.test.ts
Agent 2 → Fix batch-completion-behavior.test.ts
Agent 3 → Fix tool-approval-race-conditions.test.ts

⏸️ REQUEST APPROVAL: "May I dispatch these 3 agents in parallel?"

After approval - Work:

• Agent 1: Replaced timeouts with event-based waiting
• Agent 2: Fixed event structure bug (threadId in wrong place)
• Agent 3: Added wait for async tool execution to complete

⏸️ REPORT SUMMARIES: "All agents returned:

• Agent 1: Replaced setTimeout with event-based waiting
• Agent 2: Fixed threadId placement in event structure
• Agent 3: Added async wait for tool execution

Ready to integrate and test?"

After approval - Integration: All fixes independent, no conflicts, full suite green

⏸️ REPORT RESULTS: "Integration complete. All tests pass. 6 failures resolved."

Time saved: 3 problems solved in parallel vs sequentially

Key Benefits

1. Parallelization - Multiple investigations happen simultaneously
2. Focus - Each agent has narrow scope, less context to track
3. Independence - Agents don't interfere with each other
4. Speed - 3 problems solved in time of 1
5. Safety - Approval gates ensure user awareness at each phase

Verification

After agents return:

1. Review each summary - Understand what changed
2. Check for conflicts - Did agents edit same code?
3. ⏸️ REQUEST APPROVAL - Before running full test suite
4. Run full suite - Verify all fixes work together
5. Spot check - Agents can make systematic errors
6. ⏸️ REPORT RESULTS - Show final verification output

OpenAgent Approval Gate Summary

Approval points in parallel agent dispatch:

1. ⏸️ Before identifying independent domains
2. ⏸️ Before creating agent tasks (review scope)
3. ⏸️ Before dispatching agents in parallel
4. ⏸️ After agents return (report summaries)
5. ⏸️ Before integrating changes
6. ⏸️ Before running full test suite
7. ⏸️ After verification (report results)

Why approval gates matter:

• Prevents hasty task decomposition
• Ensures user awareness of parallel work
• Aligns with OpenAgent's safety-first philosophy
• Creates natural checkpoints for review
• Prevents integration conflicts

Red Flags - STOP and Ask

• Dispatching agents without domain analysis
• No clear boundaries between agent tasks
• Agents working on related/dependent problems
• Skipping approval gates to "move faster"
• Integrating changes without user review
• "I'll just dispatch these quickly"
• "User doesn't need to know about internal parallelization"

All of these mean: Request approval before proceeding.

Real-World Impact

From debugging session (2025-10-03):

• 6 failures across 3 files
• 3 agents dispatched in parallel
• All investigations completed concurrently
• All fixes integrated successfully
• Zero conflicts between agent changes
• User maintained awareness at each phase

Final Rule

Parallel agent work → approval before dispatch → approval before integration
Otherwise → not OpenAgent parallel dispatch

No exceptions without your human partner's permission.