Dispatching Parallel Agents
Split several independent problems across isolated sub-agents that work at the same time
- Repository
- obra/superpowers
- Stars
- 310,931
- Licence
- MIT
- Files
- 1 file
What it does
When an AI agent faces several unrelated problems, such as multiple failing test files in different parts of a codebase, this skill teaches it to hand each problem to its own sub-agent with carefully written, self-contained instructions. Those sub-agents then work at the same time instead of one after another, and the main agent keeps its own context free for coordinating and checking their results. The skill also explains when not to do this, for example when failures are related or when agents would edit the same files.
Great for
- Developers who have several unrelated test failures or bugs in different modules and want them worked on faster
- Teams using agents that support sub-agents and want a clear pattern for splitting work safely
- People learning how to write focused, well-scoped instructions for delegated AI agents
Skip it if
- Tasks where failures share a root cause, since separate agents may fix the same thing twice or conflict (the skill itself warns to investigate related failures together first)
- Tasks where tasks edit the same files or share resources, since agents can interfere with each other and cause conflicts
- Exploratory debugging where you do not yet know what is broken, because the skill assumes the problems are already identified
Try asking
- “There are 3 test files failing for different reasons after my refactor, so split them across agents and fix them in parallel”
- “Dispatch one agent per broken subsystem and tell each one to return a summary of root cause and changes”
- “Help me write focused sub-agent prompts for these five independent bugs with clear constraints on what not to change”
Install in 3 steps
- 1
Download the skill
The ZIP contains the 1 original files, the MIT licence and an attribution note.
dispatching-parallel-agents.zip - 2
Add it to your agent
Claude Code (all projects)
mkdir -p ~/.claude/skills && unzip dispatching-parallel-agents.zip -d ~/.claude/skills/Claude Code / Agent SDK (this project only)
mkdir -p .claude/skills && unzip dispatching-parallel-agents.zip -d .claude/skills/Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.
- 3
Ask for the task
No special command needed — the agent reads the skill's description and loads it when your request matches. Try one of the prompts above.
Get better results
- Only use it when you can clearly confirm the problems are independent; ask the agent to check for shared files before dispatching
- Give each sub-agent specific scope, the exact error messages, and a clear output format, and state what it must not change
- After agents return, review their changes for conflicts and run the full test suite, since sub-agents can make systematic mistakes
What's inside
- SKILL.md
Redistributed unmodified from obra/superpowers under the MIT licence. The assessment was drafted with AI from the skill's files on 2026-10-29. Skills can instruct your agent to run code — read SKILL.md and any scripts before installing.