oh-my-claudecode/seminar/quickref.md

oh-my-claudecode Quick Reference Card

v3.6.3 | github.com/Yeachan-Heo/oh-my-claudecode

Install

/plugin marketplace add https://github.com/Yeachan-Heo/oh-my-claudecode
/plugin install oh-my-claudecode
/oh-my-claudecode:omc-setup

Execution Modes

Mode	Keyword	Use Case	Example
Autopilot	autopilot	Full autonomous build	autopilot: build a REST API
Ultrapilot	ultrapilot	Parallel autopilot (3-5x)	ultrapilot: build dashboard
Ultrawork	ulw	Parallel task fixing	ulw fix all errors
Ecomode	eco	Budget-friendly parallel	eco: implement feature
Swarm	swarm	N coordinated agents	/swarm 5:executor "fix errors"
Pipeline	pipeline	Sequential chaining	/pipeline review
Ralph	ralph	Persistence until done	ralph: refactor auth
Plan	plan	Planning interview	plan the API design

Combine Modes

ralph ulw: migrate database = persistence + parallelism

Agent Tiers (28 Total)

Domain	Haiku (fast)	Sonnet (balanced)	Opus (complex)
Analysis	architect-low	architect-medium	architect
Execution	executor-low	executor	executor-high
Search	explore	-	explore-high
Frontend	designer-low	designer	designer-high
Testing	-	qa-tester	-
Security	security-rev-low	-	security-reviewer
Data Sci	-	scientist	scientist-high
Research	-	researcher	-
Build	-	build-fixer	-
TDD	tdd-guide-low	tdd-guide	-
Code Review	-	-	code-reviewer
Docs	writer	-	-
Visual	-	vision	-
Planning	-	-	planner
Critique	-	-	critic

Pipeline Presets

Preset	Flow
review	explore → architect → critic → executor
implement	planner → executor → tdd-guide
debug	explore → architect → build-fixer
research	parallel(researcher, explore) → architect → writer
refactor	explore → architect-medium → executor-high → qa-tester
security	explore → security-reviewer → executor → security-reviewer-low

Custom: /pipeline explore:haiku -> architect:opus -> executor:sonnet

Key Commands

Command	Purpose
/oh-my-claudecode:omc-setup	Initial setup wizard
/oh-my-claudecode:hud setup	Enable HUD statusline
/oh-my-claudecode:omc-doctor	Diagnose issues
/oh-my-claudecode:omc-help	Show usage guide
/oh-my-claudecode:cancel	Stop current operation
/oh-my-claudecode:note	Save compaction-resilient note
/oh-my-claudecode:learner	Extract reusable skill
/oh-my-claudecode:analyze	Deep analysis/debugging
/oh-my-claudecode:deepsearch	Thorough codebase search
/oh-my-claudecode:ultraqa	QA cycling (test/fix/repeat)
/oh-my-claudecode:tdd	Test-driven development mode

Natural Language (No Commands Needed)

• "build me a todo app" → Autopilot activates
• "fix all errors fast" → Ultrawork activates (or config default)
• "don't stop until done" → Ralph activates
• "plan the authentication" → Planning interview starts
• "stop" / "cancel" → Intelligently cancels active operation

Delegation Categories (Auto-Detection)

Category	Model	Temp	Thinking	Use For
visual-engineering	Opus	0.7	high	UI/UX, frontend, design
ultrabrain	Opus	0.3	max	Complex reasoning, architecture
artistry	Sonnet	0.9	medium	Creative solutions
quick	Haiku	0.1	low	Simple lookups
writing	Sonnet	0.5	medium	Documentation

Plan Notepads (Wisdom Capture)

Location: .omc/notepads/{plan-name}/

• learnings.md - Technical discoveries and patterns
• decisions.md - Architectural and design decisions
• issues.md - Known issues and workarounds
• problems.md - Blockers and challenges

State Files

• .omc/state/ultrapilot-state.json - Ultrapilot session
• .omc/state/ultrapilot-ownership.json - File ownership
• .omc/state/swarm-{id}.json - Swarm coordination
• .omc/state/pipeline-{id}.json - Pipeline progress

Configuration

File: ~/.claude/.omc-config.json

{
  "defaultExecutionMode": "ultrawork",  // or ""
  "maxParallelAgents": 5,
  "verificationEnabled": true
}

Verification Protocol (Built-in)

Before claiming completion:

1. IDENTIFY - What command proves this?
2. RUN - Execute verification
3. READ - Check output for pass/fail
4. CLAIM - Only then say "done" with evidence

Standard Checks: BUILD, TEST, LINT, FUNCTIONALITY, ARCHITECT, TODO, ERROR_FREE

Tips

• Combine modes: ralph ulw, ralph eco, ralplan (ralph + plan)
• Explicit keywords override defaults: eco beats config, ulw beats config
• Conflict resolution: Both ulw and eco → eco wins (more restrictive)
• Generic "fast"/"parallel" → Uses config defaultExecutionMode (default: ultrawork)
• State cleanup: /cancel --all clears all states
• Resume background: Use resume-session tool for interrupted agents
• LSP diagnostics: Full project type checking with lsp_diagnostics_directory

Resources

• GitHub: github.com/Yeachan-Heo/oh-my-claudecode
• Docs: /docs/REFERENCE.md
• Website: yeachan-heo.github.io/oh-my-claudecode-website
• NPM: npm i -g oh-my-claudecode
• Discord: (community support - link in GitHub)

---

Pro Tips:

• Start with autopilot for new projects - it handles everything
• Use ultrapilot when you need speed (3-5x faster, parallel workers)
• Use ralph when you absolutely need completion guarantee
• Use eco when managing token budgets on large tasks
• Use swarm for distributed work across many files
• Use pipeline for multi-stage workflows with quality gates