dispatching-parallel-agents Skill: Fix Independent Problems at Once

Makes your agent split two or more independent problems across parallel subagents, then review and merge what they return.

Installs: 224.6K; on skills.sh
Stars: 310.6K; on GitHub
Updated: Oct 2026; last repo commit

The skill triggers when there are 2 or more tasks with no shared state and no order between them, such as three test files failing for different reasons. It tells the agent to group the problems by area, write one focused prompt per area, send all subagents in the same response so they run at once, then read each summary, check for conflicts and run the full test suite.

Each subagent gets only the context it needs, a clear goal, constraints such as "do not change other code", and an expected output. The skill says not to use it when failures are related, when you do not yet know what is broken, or when agents would edit the same files.

Good for

Developers facing several failing test files with different causes
Claude Code and Codex users who want to keep the main session's context clean

Install dispatching-parallel-agents

Claude Code

Copy this prompt.

Install the dispatching-parallel-agents skill in Claude Code. Run this command:

npx skills add https://github.com/obra/superpowers --skill dispatching-parallel-agents

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js.

From its listing on skills.sh. Source.

What's in the skill

When it loads

Use when facing 2+ independent tasks that can be worked on without shared state or sequential dependencies

Sections in SKILL.md

Overview
When to Use
The Pattern
1. Identify Independent Domains
2. Create Focused Agent Tasks
3. Dispatch in Parallel
4. Review and Integrate
Agent Prompt Structure
Common Mistakes
When NOT to Use
Real Example from Session
Verification

Key facts

Author: obra (Superpowers)
License: MIT, from the crawl of the repo
Installs on skills.sh: 224,609 (Nov 2, 2026)
GitHub stars: 310,635 for the obra/superpowers repo (Nov 2, 2026)
Last repo commit: Oct 29, 2026
What installs: One SKILL.md; no bundled scripts
Needs: An agent client that can start subagents

What it includes

Groups failures by what is broken, such as one test file per subsystem
Fixing one domain must not affect the others.
Read · 1. Identify independent domains
Gives each agent a scope, a goal, constraints and an expected output
2. Create focused agent tasks
Sends all subagent calls in one response so they run concurrently
One call per response means they run in sequence.
Runs code · 3. Dispatch in parallel
Reads summaries, checks for conflicts, runs the full test suite
The skill also says to spot check, since agents can make systematic errors.
Runs code · 4. Review and integrate
Skip it for related failures, unknown problems or shared files
When not to use

From the official docs

Permissions

Low risk

Several subagents may edit your code at the same time. Work on a branch and review the combined diff.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

Parallel agents can edit the same code
The skill says to check for conflicts and run the full test suite afterwards. Work on a branch and read the final diff.
Agents can make systematic errors
Spot check what each agent changed instead of trusting the summaries.

Questions

What does dispatching-parallel-agents do?

It teaches your agent to hand independent problems to separate subagents at the same time instead of fixing them one after another.

How do I install it in Claude Code, the Claude app or Codex?

Use the install box above for this skill alone. The Superpowers README also lists a Claude Code plugin (/plugin install superpowers@claude-plugins-official) and Codex plugin routes. The agent must be able to start subagents.

Does it need an API key?

SKILL.md mentions no key.

Is it safe?

The skill bundles no scripts. The subagents it starts can edit files and run tests in parallel, so use version control and read each summary and the final diff.

When should I not use it?

When failures are related, when you do not know what is broken yet, when the whole system state matters, or when agents would edit the same files.

Is it meant for a single bug?

No. The description says to use it for 2 or more independent tasks. The SKILL.md gives the example of 6 test failures across 3 files handled by 3 agents with no conflicts.

Information

updated Nov 2, 2026

Repository: obra/superpowers
Folder: skills/dispatching-parallel-agents

obra/superpowers skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 2

obra/superpowers dispatching-parallel-agents SKILL.md · github.com
obra/superpowers README · github.com