Now liveThe Skillselion MCP - thousands of ranked skills, loaded into your agent mid-task. No install.Get it →
drdanmaggs avatar

Rocket Fuel

  • Updated July 25, 2026
  • drdanmaggs/rocket-fuel

rocket-fuel is a Claude Code skill in the Code Review & Quality category. Multi-agent orchestrator for Claude Code — skills, agents, and rules for TDD, code review, shipping, and project management

Key points

  • rocket-fuel
  • Code Review & Quality
  • AI-coding skill

Rocket Fuel by the numbers

  • Data as of Jul 26, 2026 (Skillselion catalog sync)
/plugin marketplace add drdanmaggs/rocket-fuel
/plugin install rocket-fuel@rocket-fuel

Add your badge

Show developers this skill is listed on Skillselion. Paste this into your README.

Listed on Skillselion
Last updatedJuly 25, 2026
Repositorydrdanmaggs/rocket-fuel

What it does

Multi-agent orchestrator for Claude Code — skills, agents, and rules for TDD, code review, shipping, and project management

README.md

Rocket Fuel

A multi-agent orchestrator that multiplies developer productivity by composing existing tools — Claude Code, GitHub Projects, tmux-CC/iTerm2, and git worktrees.

Based on the Visionary/Integrator model from Rocket Fuel by Gino Wickman & Mark C. Winters.

The idea

You're the Visionary — ideas, direction, product thinking. You talk to the Integrator, an AI agent that manages execution: spawning workers, tracking progress, protecting the current sprint from your destabilising brilliance.

Workers are ephemeral Claude Code instances running in isolated git worktrees, each picking up a GitHub issue and delivering a PR using your existing skills (/tdd, /bug-fix, /epc).

The Integrator never says no. It scopes your ideas, parks them in Someday/Maybe, and redirects back to the current epic.

Read docs/vision.md for the full concept.

Prerequisites

  • tmuxbrew install tmux
  • iTerm2 — for tmux -CC control mode (native tabs per agent)
  • Claude Codenpm install -g @anthropic-ai/claude-code
  • GitHub CLIbrew install gh
  • Gobrew install go (for building from source)

Install

# From source
git clone https://github.com/drdanmaggs/rocket-fuel.git
cd rocket-fuel
make install

# Verify
rocket-fuel version

Quick start

1. Set up a GitHub Project

Create a GitHub Project for your repo with these columns:

Someday/MaybeBacklogScopedIn ProgressReviewDone

2. Link your project

cd your-repo
rocket-fuel project link https://github.com/users/yourname/projects/1

3. Start the session

rocket-fuel up

This creates a tmux session with Integrator and Dashboard tabs, then attaches with tmux -CC. iTerm2 renders each as a native tab.

4. Spawn a worker

Label an issue with a workflow (e.g., workflow:tdd) and spawn a worker:

rocket-fuel work 42
# or
rocket-fuel work https://github.com/yourname/repo/issues/42

This creates a git worktree, opens a new iTerm2 tab, and launches Claude Code with the issue context and appropriate skill.

5. Monitor

# One-shot status
rocket-fuel status

# Live dashboard (in the dashboard tab)
watch -n 30 rocket-fuel status

# View the project board
rocket-fuel project board

6. Surface a tab

The Integrator brings tabs to the foreground when the Visionary's attention is needed:

rocket-fuel surface worker-42 "CI failing — needs your eyes"

This switches the active iTerm2 tab and sends a macOS notification.

7. Clean up

# Remove completed workers (worktrees + windows)
rocket-fuel reap

# Tear down the entire session
rocket-fuel down

Skill routing

Issues are routed to skills based on labels:

Label Skill Approach
workflow:tdd /tdd RED → GREEN → REFACTOR
workflow:bug-fix /bug-fix Failing test first
workflow:epc /epc Explore → Plan → Code
workflow:issue-scope /issue-scope Break down into sub-issues
(no label) /epc Default

Development

make setup          # Configure git hooks
make build          # Build binary
make test           # Run tests (with race detector)
make test-integration  # Run integration tests (needs tmux)
make lint           # golangci-lint
make fmt            # Format code
make all            # fmt + lint + test + build

Prior art

Inspired by gastown (Steve Yegge's multi-agent workspace manager) but philosophically opposite — compose existing tools rather than building a platform. See docs/gastown-lessons.md.

Related skills

This week in AI coding

Five minutes, every Monday - the tools, releases and tactics for developers.

unsubscribe anytime.