Skip to content

Skill improvements: skills__dispatching-parallel-agents #2

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
RichardHightower wants to merge 3 commits into
base: main
Choose a base branch
from
Open

Skill improvements: skills__dispatching-parallel-agents #2

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
3 commits
Select commit Hold shift + click to select a range
0ae604e
Skill improvements: skills__dispatching-parallel-agents
RichardHightower Dec 30, 2025
2c362e8
Skill improvements: skills__dispatching-parallel-agents
RichardHightower Dec 30, 2025
1c61476
Skill improvements: skills__dispatching-parallel-agents
RichardHightower Dec 30, 2025
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
133 changes: 49 additions & 84 deletions skills/dispatching-parallel-agents/SKILL.md
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -5,12 +5,36 @@ description: Use when facing 2+ independent tasks that can be worked on without

# Dispatching Parallel Agents

## Table of Contents

1. [Overview](#overview)
2. [Quick Reference Checklist](#quick-reference-checklist)
3. [When to Use](#when-to-use)
4. [The Pattern](#the-pattern)
5. [Agent Prompt Structure](#agent-prompt-structure)
6. [Common Mistakes](#common-mistakes)
7. [Verification](#verification)
8. [Key Benefits](#key-benefits)
9. [Reference Files](#reference-files)

## 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.
Multiple unrelated failures (different test files, different subsystems, different bugs) waste time when investigated sequentially. Each investigation is independent and can happen in parallel.

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

## Quick Reference Checklist

Use this checklist for rapid parallel dispatch:

1. ☐ **Group failures by domain** - Identify independent subsystems
2. ☐ **Verify independence** - Confirm no shared state between problems
3. ☐ **Create focused prompts** - One clear scope per agent ([template](references/agent-prompt-template.md))
4. ☐ **Dispatch in parallel** - Launch all agents simultaneously
5. ☐ **Review summaries** - Read what each agent found
6. ☐ **Check for conflicts** - Verify no overlapping changes
7. ☐ **Run full suite** - Validate all fixes work together

## When to Use

```dot
Expand All @@ -37,10 +61,11 @@ digraph when_to_use {
- 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
**Avoid when:**
- Failures are related (fixing one might fix others)
- Full system state understanding required first
- Unknown failure cause - investigate before dispatching
- Agents would interfere (editing same files, using same resources)

## The Pattern

Expand All @@ -59,7 +84,7 @@ 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
- **Expected output:** Summary of what found and fixed

### 3. Dispatch in Parallel

Expand All @@ -86,74 +111,24 @@ Good agent prompts are:
2. **Self-contained** - All context needed to understand the problem
3. **Specific about output** - What should the agent return?

```markdown
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.
```
See the complete reusable template: **[Agent Prompt Template](references/agent-prompt-template.md)**

## 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"

## 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
| Anti-Pattern | Problem | Fix |
|-------------|---------|-----|
| "Fix all the tests" | Too broad, agent gets lost | Scope to one file/subsystem |
| "Fix the race condition" | No location context | Paste error messages and test names |
| No constraints | Agent might refactor everything | Add "Do NOT change production code" |
| "Fix it" | Vague output expectation | "Return summary of root cause and changes" |

**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
```

**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

**Integration:** All fixes independent, no conflicts, full suite green
## Verification

**Time saved:** 3 problems solved in parallel vs sequentially
After agents return:
1. **Review each summary** - Understand what changed
2. **Check for conflicts** - Did agents edit same code?
3. **Run full suite** - Verify all fixes work together
4. **Spot check** - Agents can make systematic errors

## Key Benefits

Expand All @@ -162,19 +137,9 @@ Agent 3 → Fix tool-approval-race-conditions.test.ts
3. **Independence** - Agents don't interfere with each other
4. **Speed** - 3 problems solved in time of 1

## Verification

After agents return:
1. **Review each summary** - Understand what changed
2. **Check for conflicts** - Did agents edit same code?
3. **Run full suite** - Verify all fixes work together
4. **Spot check** - Agents can make systematic errors

## Real-World Impact
## Reference Files

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
| File | Description |
|------|-------------|
| [Agent Prompt Template](references/agent-prompt-template.md) | Reusable template with placeholder variables |
| [Example Session](references/example-session.md) | Real debugging session demonstrating the pattern |
79 changes: 79 additions & 0 deletions skills/dispatching-parallel-agents/references/agent-prompt-template.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,79 @@
# Agent Prompt Template

Use this template when dispatching parallel agents for independent problem domains.

## Template Structure

```markdown
Fix the {{NUMBER}} failing tests in {{FILE_PATH}}:

{{NUMBERED_LIST_OF_FAILURES}}

These are {{PROBLEM_TYPE}} issues. Your task:

1. Read the test file and understand what each test verifies
2. Identify root cause - {{DIAGNOSTIC_QUESTIONS}}
3. Fix by:
- {{FIX_APPROACH_1}}
- {{FIX_APPROACH_2}}
- {{FIX_APPROACH_3}}

Do NOT {{CONSTRAINTS}}.

Return: {{EXPECTED_OUTPUT_FORMAT}}
```

## Placeholder Definitions

| Placeholder | Description | Example |
|------------|-------------|---------|
| `{{NUMBER}}` | Count of failures in scope | `3` |
| `{{FILE_PATH}}` | Relative path to test file | `src/agents/agent-tool-abort.test.ts` |
| `{{NUMBERED_LIST_OF_FAILURES}}` | Test names with error context | See below |
| `{{PROBLEM_TYPE}}` | Category of issue | `timing/race condition`, `null reference`, `async` |
| `{{DIAGNOSTIC_QUESTIONS}}` | What agent should investigate | `timing issues or actual bugs?` |
| `{{FIX_APPROACH_N}}` | Allowed fix strategies | `Replace arbitrary timeouts with event-based waiting` |
| `{{CONSTRAINTS}}` | What NOT to do | `just increase timeouts - find the real issue` |
| `{{EXPECTED_OUTPUT_FORMAT}}` | What agent returns | `Summary of what you found and what you fixed` |

## Example: Filled Template

```markdown
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.
```

## Prompt Quality Checklist

Before dispatching, verify the prompt is:

- [ ] **Focused** - Single clear problem domain
- [ ] **Self-contained** - All context needed to understand the problem
- [ ] **Specific about output** - Clear expectation for what agent returns
- [ ] **Constrained** - Boundaries on what NOT to change
- [ ] **Actionable** - Concrete steps agent can follow

## Common Anti-Patterns

| Anti-Pattern | Problem | Fix |
|-------------|---------|-----|
| "Fix all the tests" | Too broad, agent gets lost | Scope to one file/subsystem |
| "Fix the race condition" | No context about location | Paste error messages and test names |
| No constraints | Agent might refactor everything | Add "Do NOT change production code" |
| "Fix it" | Vague output expectation | "Return summary of root cause and changes" |
80 changes: 80 additions & 0 deletions skills/dispatching-parallel-agents/references/example-session.md
View file
Open in desktop
Original file line number Diff line number Diff line change
@@ -0,0 +1,80 @@
# Real-World Example: Parallel Agent Debugging Session

This example demonstrates the parallel agent pattern applied to a real debugging session (2025-10-03).

## Initial Situation

**Context:** 6 test failures across 3 files after major refactoring

**Failures identified:**
| File | Failures | Category |
|------|----------|----------|
| `agent-tool-abort.test.ts` | 3 | Timing issues |
| `batch-completion-behavior.test.ts` | 2 | Tools not executing |
| `tool-approval-race-conditions.test.ts` | 1 | Execution count = 0 |

## Decision Analysis

**Question:** Are these failures independent?

**Analysis:**
- Abort logic is separate from batch completion
- Batch completion is separate from race condition handling
- No shared state between subsystems
- Each can be investigated without context from others

**Decision:** ✅ Independent domains → Dispatch parallel agents

## Agent Dispatch

Three agents dispatched simultaneously:

```
Agent 1 → Fix agent-tool-abort.test.ts (3 failures)
Agent 2 → Fix batch-completion-behavior.test.ts (2 failures)
Agent 3 → Fix tool-approval-race-conditions.test.ts (1 failure)
```

## Agent Results

### Agent 1: Abort Tests
**Root Cause:** Arbitrary timeouts causing flaky behavior
**Fix:** Replaced timeouts with event-based waiting
**Files Changed:** `agent-tool-abort.test.ts`

### Agent 2: Batch Completion Tests
**Root Cause:** Event structure bug - `threadId` in wrong place
**Fix:** Corrected event payload structure
**Files Changed:** `batch-completion-behavior.test.ts`, `event-handler.ts`

### Agent 3: Race Condition Tests
**Root Cause:** Test not waiting for async tool execution
**Fix:** Added wait for async tool execution to complete
**Files Changed:** `tool-approval-race-conditions.test.ts`

## Integration

**Conflict Check:**
- Agent 1 touched abort test file only ✅
- Agent 2 touched batch test + event handler ✅
- Agent 3 touched race condition test only ✅
- No overlapping file changes ✅

**Full Suite Run:** All tests pass ✅

## Metrics

| Metric | Value |
|--------|-------|
| Total failures | 6 |
| Agents dispatched | 3 |
| Files fixed | 4 |
| Conflicts | 0 |
| Time multiplier | ~3x faster than sequential |

## Key Takeaways

1. **Domain separation was correct** - Fixes were truly independent
2. **No agent conflicts** - Each worked in isolated scope
3. **Parallel execution saved time** - 3 problems solved in time of 1
4. **Clear prompts produced clear results** - Each agent knew exactly what to investigate