Skip to main content
This guide gets you to a working state in under 30 seconds. For detailed installation options, see the Installation page.

Choose Your Installation Method

Verify Installation

Run a health check to ensure everything is configured correctly:
/doctor
The /doctor command checks: plugin installation, hooks status, context health, MCP servers, CLAUDE.md size, and git status.

Try Your First Command

Test your setup with the /develop command for multi-phase feature development:
1

Start a Feature

/develop add user authentication
The orchestrator agent will guide you through:
  • Research - Explore codebase, score confidence (0-100)
  • Plan - Present approach, wait for approval
  • Implement - Execute with quality gates
  • Review - Security and quality checks
2

End Your Session

/wrap-up
Runs the end-of-session checklist:
  • Changes audit (git status, git diff --stat)
  • Quality checks (lint, typecheck, tests)
  • Learning capture (what mistakes? what worked?)
  • Next session context
3

Capture a Learning

When you correct the AI:
/learn-rule
The AI proposes a rule to prevent future mistakes. You approve, and it’s added to your LEARNED section.

Core Commands Reference

/develop

Multi-phase feature build with validation gates

/wrap-up

End-of-session ritual with learning capture

/learn-rule

Extract corrections into persistent memory

/commit

Smart commit with quality gates

What’s Next?

Installation Details

Learn about all installation methods and configuration options

Core Patterns

Understand the patterns that make Pro Workflow powerful

Commands

Explore all 10 commands and their usage

Agents

Use specialized agents for planning, review, and debugging

The Core Insight

“80% of my code is written by AI, 20% is spent reviewing and correcting it.” — Karpathy
Pro Workflow optimizes for that ratio. Every pattern reduces correction cycles:
  • Self-Correction Loop - Mistakes become rules that prevent future mistakes
  • Parallel Worktrees - Zero dead time while Claude thinks
  • 80/20 Review - Batch reviews at checkpoints, not every change
  • Context Discipline - Manage your 200k token budget effectively
  • Orchestration - Wire Commands > Agents > Skills for complex features
After 50 sessions with self-correction enabled, Claude barely needs correcting. The system trains itself through your corrections.

Build docs developers (and LLMs) love