parallel-agents（平行 Agent）

Dispatching Parallel Agents

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. Each agent gets its own planning directory for structured knowledge capture.

Core principle: Dispatch one agent per independent problem domain with its own planning dir. Let them work concurrently.

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

1. Identify Independent Domains

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.

2. Create Per-Agent Planning Directories

Before dispatching, create a planning dir for each agent:

mkdir -p .planning/agents/{role}-task-{N}/

Example for 3 parallel agents:

mkdir -p .planning/agents/fixer-abort-tests/
mkdir -p .planning/agents/fixer-batch-tests/
mkdir -p .planning/agents/fixer-approval-tests/

3. Create Focused Agent Tasks

Each agent gets:

• Specific scope: One test file or subsystem
• Clear goal: Make these tests pass
• Constraints: Don't change other code
• Planning dir: Path to its .planning/agents/{role}/ directory
• Planning rules: The 6 rules for structured knowledge capture
• Expected output: Summary of what you found and fixed

4. Dispatch in Parallel

// In Claude Code / AI environment
// Each agent gets its planning dir in the prompt
Task("Fix agent-tool-abort.test.ts failures\n\nPlanning dir: .planning/agents/fixer-abort-tests/\n[planning rules]")
Task("Fix batch-completion-behavior.test.ts failures\n\nPlanning dir: .planning/agents/fixer-batch-tests/\n[planning rules]")
Task("Fix tool-approval-race-conditions.test.ts failures\n\nPlanning dir: .planning/agents/fixer-approval-tests/\n[planning rules]")
// All three run concurrently

5. Review, Aggregate, and Integrate

When all agents return:

Batch aggregation:

1. Read each agent's .planning/agents/{role}/findings.md and progress.md
2. Extract critical items, errors, and test results from each
3. Append to top-level .planning/findings.md under agent/task headings
4. Update top-level .planning/progress.md: update Task Status Dashboard table + append completion details

Integration:

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

Example aggregation:

<!-- Append to .planning/findings.md -->
## Parallel Fix Session: Test Failures

### Agent: fixer-abort-tests
- Root cause: race condition in abort handler timing
- Fix: replaced setTimeout with event-based waiting
- [From agent] > **Critical for Orchestrator:** abort handler needs upstream fix for edge case

### Agent: fixer-batch-tests
- Root cause: threadId in wrong position in event structure
- Fix: moved threadId to top-level event property

### Agent: fixer-approval-tests
- Root cause: async tool execution not awaited
- Fix: added proper await for tool completion

<!-- Append to .planning/progress.md -->
- [x] Parallel fix session: 3 agents dispatched
  - fixer-abort-tests: COMPLETED (3 tests fixed)
  - fixer-batch-tests: COMPLETED (2 tests fixed)
  - fixer-approval-tests: COMPLETED (1 test fixed)
  - Integration: All fixes independent, no conflicts

Agent Prompt Structure with Planning

Good agent prompts include planning dir injection:

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.

## Planning Directory

Your planning directory is: .planning/agents/fixer-abort-tests/

You MUST follow the 6 planning rules from planning-foundation/templates/agent-context.md

Mark critical items with: `> **Critical for Orchestrator:** [description]`

## 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"

No planning dir: Agent's findings get lost after it exits With planning dir: Knowledge persists for orchestrator aggregation

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 from Session

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)

Decision: Independent domains - abort logic separate from batch completion separate from race conditions

Setup:

mkdir -p .planning/agents/fixer-abort-tests/
mkdir -p .planning/agents/fixer-batch-tests/
mkdir -p .planning/agents/fixer-approval-tests/

Dispatch:

Agent 1 → Fix agent-tool-abort.test.ts (planning: .planning/agents/fixer-abort-tests/)
Agent 2 → Fix batch-completion-behavior.test.ts (planning: .planning/agents/fixer-batch-tests/)
Agent 3 → Fix tool-approval-race-conditions.test.ts (planning: .planning/agents/fixer-approval-tests/)

Results:

• 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

Aggregation: Read each agent's findings.md, compiled into.planning/findings.md

Integration: All fixes independent, no conflicts, full suite green

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. Knowledge capture - Per-agent planning dirs preserve findings for orchestrator

Verification

After agents return:

1. Aggregate findings - Read each agent's planning dir, compile into top-level.planning/
2. Review each summary - Understand what changed
3. Check for conflicts - Did agents edit same code?
4. Run full suite - Verify all fixes work together
5. Spot check - Agents can make systematic errors

Integration with Other Skills

Works well with:

• superpower-planning:subagent-driven - For sequential task execution with reviews
• superpower-planning:debugging - When parallel investigation is needed
• superpower-planning:verification - Run after integrating all agent changes

Real-World Impact

From debugging session:

• 6 failures across 3 files
• 3 agents dispatched in parallel
• All investigations completed concurrently
• All fixes integrated successfully
• Zero conflicts between agent changes
• All findings preserved in.planning/ for future reference