Claude-skill-registry autopilot

Full autonomous execution from idea to working code

install
source · Clone the upstream repo
git clone https://github.com/majiayu000/claude-skill-registry
Claude Code · Install into ~/.claude/skills/
T=$(mktemp -d) && git clone --depth=1 https://github.com/majiayu000/claude-skill-registry "$T" && mkdir -p ~/.claude/skills && cp -r "$T/skills/data/autopilot" ~/.claude/skills/majiayu000-claude-skill-registry-autopilot-661f02 && rm -rf "$T"
manifest: skills/data/autopilot/SKILL.md
source content

Autopilot Skill

Full autonomous execution from idea to working code.

Overview

Autopilot is the ultimate hands-off mode. Give it a brief product idea (2-3 lines) and it handles everything:

  1. Understands your requirements (Analyst)
  2. Designs the technical approach (Architect)
  3. Plans the implementation (Critic-validated)
  4. Builds with parallel agents (Ralph + Ultrawork)
  5. Tests until everything passes (UltraQA)
  6. Validates quality and security (Multi-architect review)

Usage

/oh-my-claudecode:autopilot <your idea>
/oh-my-claudecode:ap "A CLI tool that tracks daily habits"
/oh-my-claudecode:autopilot Add dark mode to the app

Magic Keywords

These phrases auto-activate autopilot:

  • "autopilot", "auto pilot", "autonomous"
  • "build me", "create me", "make me"
  • "full auto", "handle it all"
  • "I want a/an..."

Phases

Phase 0: Expansion

Goal: Turn vague idea into detailed spec

Agents:

  • Analyst (Opus) - Extract requirements
  • Architect (Opus) - Technical specification

Output:

.omc/autopilot/spec.md

Phase 1: Planning

Goal: Create implementation plan from spec

Agents:

  • Architect (Opus) - Create plan (direct mode, no interview)
  • Critic (Opus) - Validate plan

Output:

.omc/plans/autopilot-impl.md

Phase 2: Execution

Goal: Implement the plan

Mode: Ralph + Ultrawork (persistence + parallelism)

Agents:

  • Executor-low (Haiku) - Simple tasks
  • Executor (Sonnet) - Standard tasks
  • Executor-high (Opus) - Complex tasks

Phase 3: QA

Goal: All tests pass

Mode: UltraQA

Cycle:

  1. Build
  2. Lint
  3. Test
  4. Fix failures
  5. Repeat (max 5 cycles)

Phase 4: Validation

Goal: Multi-perspective approval

Agents (parallel):

  • Architect - Functional completeness
  • Security-reviewer - Vulnerability check
  • Code-reviewer - Quality review

Rule: All must APPROVE or issues get fixed and re-validated.

Configuration

Optional settings in

.claude/settings.json
:

{
  "omc": {
    "autopilot": {
      "maxIterations": 10,
      "maxQaCycles": 5,
      "maxValidationRounds": 3,
      "pauseAfterExpansion": false,
      "pauseAfterPlanning": false,
      "skipQa": false,
      "skipValidation": false
    }
  }
}

Cancellation

/oh-my-claudecode:cancel

Or say: "stop", "cancel", "abort"

Progress is preserved for resume.

Resume

If autopilot was cancelled or failed, just run

/oh-my-claudecode:autopilot
again to resume from where it stopped.

Examples

New Project:

/oh-my-claudecode:autopilot A REST API for a bookstore inventory with CRUD operations

Feature Addition:

/oh-my-claudecode:autopilot Add user authentication with JWT tokens

Enhancement:

/oh-my-claudecode:ap Add dark mode support with system preference detection

Best Practices

  1. Be specific about the domain - "bookstore" not "store"
  2. Mention key features - "with CRUD", "with authentication"
  3. Specify constraints - "using TypeScript", "with PostgreSQL"
  4. Let it run - Don't interrupt unless truly needed

STATE CLEANUP ON COMPLETION

IMPORTANT: Delete ALL state files on successful completion

When autopilot reaches the

complete
phase (all validation passed):

# Delete autopilot and all sub-mode state files
rm -f .omc/state/autopilot-state.json
rm -f .omc/state/ralph-state.json
rm -f .omc/state/ultrawork-state.json
rm -f .omc/state/ultraqa-state.json
rm -f ~/.claude/ralph-state.json
rm -f ~/.claude/ultrawork-state.json

This ensures clean state for future sessions.

Troubleshooting

Stuck in a phase?

  • Check TODO list for blocked tasks
  • Review
    .omc/autopilot-state.json
    for state
  • Cancel and resume if needed

Validation keeps failing?

  • Review the specific issues
  • Consider if requirements were too vague
  • Cancel and provide more detail

QA cycles exhausted?

  • Same error 3 times = fundamental issue
  • Review the error pattern
  • May need manual intervention