Skip to content
Get packs

Subagent-Driven Development

Runs a multi-task plan by delegating each task to fresh AI subagents with automatic review

Source Download skill
Repository
obra/superpowers
Stars
310,931
Licence
MIT
Files
11 files

What it does

When you hand the agent a multi-step implementation plan, this skill makes it split the work into tasks and send each task to a separate, fresh sub-agent with only the context it needs. After each task, a reviewer checks spec compliance and code quality, and a broader review runs at the end. The agent keeps a progress ledger file so it can resume after its memory is compacted, and it is told to keep going without pausing for check-ins except in a few named cases such as destructive operations, pushes, or publishes.

Great for

  • Developers with a written plan of mostly independent coding tasks who want it executed hands-off
  • Teams using git worktrees who want a per-plan ledger and review trail
  • People who want to avoid the agent losing track of progress in long sessions

Skip it if

  • Tightly coupled work where tasks depend heavily on each other, since the skill itself says to brainstorm or execute manually instead
  • Non-technical users or tasks without a plan file, since the skill is built around executing plans in a git repository
  • Environments without a subagent tool, where the skill directs you to the inline executing-plans skill instead

Try asking

  • “Execute the implementation plan in docs/plan.md, one fresh subagent per task with reviews”
  • “Work through all remaining tasks in this plan and give me a final review at the end”
  • “Resume the plan from my ledger after the session was compacted”
Step-by-step guide: how to use Subagent-Driven DevelopmentInstall, first run, examples, tips and FAQs

Install in 3 steps

  1. 1

    Download the skill

    The ZIP contains the 11 original files, the MIT licence and an attribution note.

    subagent-driven-development.zip
  2. 2

    Add it to your agent

    Claude Code (all projects)

    mkdir -p ~/.claude/skills && unzip subagent-driven-development.zip -d ~/.claude/skills/

    Claude Code / Agent SDK (this project only)

    mkdir -p .claude/skills && unzip subagent-driven-development.zip -d .claude/skills/

    Claude apps (web/desktop): Settings → Capabilities → Skills → Upload skill, and choose the ZIP.

  3. 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

  • Give it a clear plan file with independent tasks, since tightly coupled tasks are a poor fit
  • Review the ledger and workspace directory under .superpowers/sdd/ if you want to audit what each subagent did
  • Check the final review output before accepting the finish step, since the agent may merge or push only after your explicit consent

What's inside

  • implementer-prompt.md
  • re-review-prompt.md
  • references/batching.md
  • references/cannot-verify.md
  • references/final-review.md
  • references/fix-loop.md
  • references/handling-reports.md
  • references/preflight.md
  • references/resuming.md
  • SKILL.md
  • task-reviewer-prompt.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.