home / AI Agents & MCP / Subagent Driven Development Skill by @obra

Subagent Driven Development Skill by @obra

FREE

github stars★ 3,07,678updatedOct 2026repo@obra

activity (commits, last 6 months)

6mo ago1

this month2

$git clone https://github.com/obra/superpowers

The skill is the skills/subagent-driven-development folder inside that repository. Copy it into the folder your agent loads skills from; check your agent's documentation for where that is.

about Subagent Driven Development Skill by @obra

Use when executing implementation plans with independent tasks in the current session.

our take

Executes an implementation plan by handing independent tasks to subagents within the current session, with notes on model selection and a task loop. It is the hands-on counterpart to plan writing.

when to use it

You have a written plan with independent tasks
You want fresh context for each task instead of one long session
You want to choose lighter or stronger models per task

before you install

Needs a clear plan first, or the subagents will interpret tasks differently
Check each task's output before the next begins

what's inside

The original SKILL.md is organised into these sections:

When to Use
The Process
Setup
Model Selection
The Task Loop
1. Dispatch the implementer
2. Handle the report
3. Review the task
4. The fix loop
5. Complete the task
Final Review
Common Rationalizations

source and license

Written and maintained by @obra. It is published under the MIT license, which allows reuse with attribution. SkillDirectory does not claim authorship: this page adds our own review and links back to the original, and the skill's text belongs to its author.

View the full SKILL.md on GitHub ↗

Read the opening of the original SKILL.md

# Subagent-Driven Development

Execute plan by dispatching a fresh implementer subagent per task, a task review (spec compliance + code quality) after each, and a broad whole-branch review at the end.

**Why subagents:** You delegate tasks to specialized agents with isolated context. By precisely crafting their instructions and context, you ensure they stay focused and succeed at their task. They should never inherit your session's context or history — you construct exactly what they need. This also preserves your own context for coordination work.

**Core principle:** Fresh subagent per task + task review (spec + quality) + broad final review = high quality, fast iteration

**Narration:** between tool calls, narrate at most one short line — the
ledger and the tool results carry the record.
…

Excerpt only, quoted from the author's repository. Open the repository for the complete file.

↓ download .zip