Oh My Pilot 🚀

Multi-agent orchestration plugin for OpenCode with 32 specialized agents, 40 intelligent skills, and 5 powerful tool categories.

✨ Features

• 32 Specialized Agents - Expert AI agents across 4 categories (advisor, executor, creator, specialist)
• 40 Intelligent Skills - Composable behaviors for autopilot, ultrapilot, planning, UI/UX, Git mastery, and more
• 5 Tool Categories - AST (code search/replace), LSP (code intelligence), Python REPL, Browser Automation, Spreadsheet Tools
• Magic Keywords - Natural language activation (autopilot, ultrapilot, ultrawork, planner, ralph, ecomode)
• Silent Skills - Context-aware activation (frontend-ui-ux, git-master)
• Intelligent Routing - Automatic agent and skill selection based on intent
• Analytics & Tracking - Token usage, cost estimation, session management

Installation

Add to your opencode.json:

{
  "plugin": ["oh-my-pilot"]
}

Or install locally:

cd packages/oh-my-pilot
bun install
bun run build

Quick Start

Agent Delegation

Delegate tasks to specialized AI agents:

# Use the architect agent for design decisions
pilot.architect: "Design a REST API for user management"

# Use the executor agent for implementation
pilot.executor: "Implement the user registration endpoint"

# Use the researcher agent to explore code
pilot.researcher: "Find all files that handle authentication"

Magic Keywords

Quick Start Guide

1. Delegate to Specialists

Use specific agents for focused work:

# Architecture design
pilot.architect: "Design microservices for e-commerce"

# Code implementation  
pilot.executor: "Implement user authentication with JWT"

# Code analysis
pilot.researcher: "Map the authentication flow"

# Code review
pilot.code-reviewer: "Review the payment module for quality"

# Security audit
pilot.security-reviewer: "Audit the API endpoints"

# Performance optimization
pilot.performance-optimizer: "Optimize the database queries"

2. Use Skills for Workflow Control

Activate with magic keywords:

# Autonomous execution (5 phases)
autopilot: Build a REST API with authentication and tests

# Parallel autopilot (3-5x faster) ⚡ NEW
ultrapilot: Build a complete e-commerce system

# Parallel agents (fast)
ultrawork: Refactor 10 API endpoints in parallel

# Planning help
planner: Help me plan a React app migration

# Persist until done
ralph: Fix all TypeScript errors. Tests must pass.

# Budget friendly
ecomode: Add documentation to all functions

Silent Skills (Context-Aware)

These activate automatically based on context:

# Frontend-UI-UX - Automatically activates for UI work 🎨 NEW
"Create a responsive navigation component"
"Design a user settings modal"

# Git-Master - Automatically activates for Git operations 🔀 NEW
"Commit these changes with conventional format"
"Push feature branch to remote"

3. Leverage Specialized Tools

# Code search with patterns
ast.search: Find all console.log calls

# Code replacement
ast.replace: Change logger.debug to logger.info

# Python scripting
python.execute: Process CSV and generate report

# Browser automation
browser.screenshot: Capture homepage at 1920x1080

# Spreadsheet operations  
spreadsheet.read: Load financial data from Excel

Documentation & Examples

• EXAMPLES.md - Complete usage examples for all features
• SKILLS_DEEP_DIVE.md - Detailed skill guide with use cases
• AGENTS_USAGE_GUIDE.md - How to use each agent effectively
• WORKFLOWS_BEST_PRACTICES.md - Patterns, tips, and troubleshooting

Complete Feature List

32 Agents

Advisor Agents (5): architect, architect-medium, architect-low, designer, planner

Executor Agents (5): executor, executor-high, executor-low, build-fixer, refactorer

Creator Agents (2): writer, documenter

Specialist Agents (20): researcher, explore, critic, analyst, qa-tester, scientist, vision, code-reviewer, security-reviewer, debugger, typescript-specialist, performance-optimizer, react-specialist, api-specialist, db-specialist, devops-specialist, testing-specialist, accessibility-specialist

31 Skills

Core (5): autopilot, ultrawork, planner, ralph, default

Extended (26): ecomode, ultraqa, swarm, ultrapilot, pipeline, tdd, analyze, research, review, security-review, optimize, document, debug, refactor, build-fix, migration, accessibility, performance, testing, deployment, monitoring, and 11 more specialized skills

5 Tool Categories

1. AST Tools - Code search, pattern matching, refactoring
2. LSP Tools - Code intelligence, diagnostics, completion
3. Python REPL - Script execution, data processing
4. Browser Automation - Capture, testing, scraping (requires puppeteer)
5. Spreadsheet Tools - Excel operations, analysis, reporting (requires xlsx)

Architecture

• OpenCode Native - Uses OpenCode's client.session.prompt() API
• Official SDK - Built with @opencode-ai/plugin and @opencode-ai/sdk for full OpenCode compliance
• Type Safe - Full TypeScript with strict mode
• Production Ready - Zero build errors, comprehensive documentation
• Scalable - 32 agents, 31 skills, 5 tool categories
• 100% Compliant - Fully conforms to OpenCode plugin standards

Installation

Add to your opencode.json:

{
  "plugin": ["oh-my-pilot"]
}

Development

# Install dependencies
npm install

# Build the plugin
npm run build

# Run tests
npm test

# Check types
npm run typecheck

Available Agents

Advisor Agents

| Agent | Description | Tier | |-------|-------------|------| | architect | Software architecture advisor | HIGH | | designer | UI/UX design advisor | MEDIUM | | planner | Strategic planning with interviews | HIGH |

Executor Agents

| Agent | Description | Tier | |-------|-------------|------| | executor | Code implementation | MEDIUM | | executor-high | Complex multi-file tasks | HIGH | | executor-low | Quick simple changes | LOW |

Specialist Agents

| Agent | Description | Tier | |-------|-------------|------| | researcher | Codebase research and analysis | MEDIUM | | explore | Fast code navigation | LOW | | code-reviewer | Comprehensive code review | HIGH | | security-reviewer | Security audit and hardening | HIGH | | performance-optimizer | Performance tuning | MEDIUM |

See AGENTS_USAGE_GUIDE.md for all 32 agents.

Skills Overview

Core Skills

| Skill | Purpose | Keywords | |-------|---------|----------| | autopilot | 5-phase autonomous execution | autopilot, ap | | ultrawork | Parallel agent execution | ultrawork, ulw | | planner | Interactive planning | plan, planner | | ralph | Persist until done | ralph, don't stop until | | ecomode | Low token usage | ecomode, eco |

See SKILLS_DEEP_DIVE.md for all 31 skills.

Tools

AST Tools

• ast.search - Search code patterns using AST matching
• ast.replace - Replace patterns while preserving matched content

LSP Tools

• lsp.diagnostics - Get errors and warnings
• lsp.hover - Get symbol information
• lsp.gotoDefinition - Find symbol definitions
• lsp.findReferences - Find all references
• lsp.documentSymbols - List file symbols
• lsp.workspaceSymbols - Search workspace symbols

Python REPL

• python.execute - Execute Python with persistent state
• python.reset - Clear session variables
• python.getState - Get current variables

Browser Automation

• browser.screenshot - Capture webpage screenshots
• browser.generatePDF - Convert to PDF
• browser.scrape - Extract data with selectors
• browser.performanceTest - Measure web metrics
• browser.visualRegression - Compare visual changes

Includes: puppeteer (pre-installed)

Spreadsheet Tools

• spreadsheet.readExcel - Read Excel files
• spreadsheet.writeExcel - Write to Excel
• spreadsheet.createPivotTable - Create pivot tables
• spreadsheet.applyFormula - Apply formulas
• spreadsheet.generateChart - Create charts
• spreadsheet.analyzeData - Statistical analysis

Requires: npm install xlsx

Analytics

Track your usage:

# Get summary of all sessions
pilot.analytics

# Get specific session report
pilot.analytics with sessionId "session-123"

# Export to CSV
pilot.analytics with format "csv"

What's New in This Release

✨ Complete Feature Set:

• 32 agents (up from 10)
• 31 skills (up from 5)
• 5 tool categories
• 100% feature parity with original plugin

🔧 Enhanced Features:

• Browser automation tools (Puppeteer integration)
• Spreadsheet processing tools (Excel support)
• Comprehensive documentation and examples

📚 Documentation:

• Complete usage examples
• Deep dive skill guides
• Agent usage instructions
• Workflow patterns and best practices

🎯 Production Ready:

• Zero build errors
• Strict TypeScript mode
• Comprehensive test suites
• Full OpenCode integration

Getting Help

For Usage Questions

See the EXAMPLES.md file for complete usage examples across all features.

For Skill Deep Dives

Check SKILLS_DEEP_DIVE.md for detailed skill explanations with real-world scenarios.

For Agent Selection

Visit AGENTS_USAGE_GUIDE.md to find the right agent for your task.

For Workflows

Explore WORKFLOWS_BEST_PRACTICES.md for patterns, tips, and troubleshooting.

For Integration

See INTEGRATION_GUIDE.md for setup and configuration details.

Architecture

This plugin is built with:

• OpenCode Native - Uses OpenCode's native client.session.prompt() API
• Official SDK Integration - Fully integrated with @opencode-ai/plugin and @opencode-ai/sdk
• Modular Design - Agent registry, skill router, tool orchestration
• Analytics Built-in - Token tracking, cost estimation, session management
• Type Safe - Full TypeScript with strict mode enabled
• OpenCode Compliant - 100% adherence to OpenCode plugin standards

Stats

| Metric | Count | |--------|-------| | Agents | 32 | | Skills | 31 | | Tool Categories | 5 | | Source Files | 21 | | Lines of Code | 5,900+ | | Test Cases | 260+ | | Documentation Pages | 8 |

License

MIT

Support

For issues or questions:

1. Check the documentation files (EXAMPLES.md, SKILLS_DEEP_DIVE.md, etc.)
2. Review the INTEGRATION_GUIDE.md for setup help
3. Check WORKFLOWS_BEST_PRACTICES.md for common issues

Happy coding! 🚀 Choose the right agent and skill for your task and let oh-my-pilot do the heavy lifting.

Analytics include:

• Token usage (input/output/total)
• Cost estimation (USD)
• Agent usage frequency
• Tool usage frequency
• Session duration

OpenCode-Native Architecture

Important: This plugin uses OpenCode's native model execution system. Model selection is controlled by your OpenCode configuration, not hardcoded in the plugin.

Model Configuration

Models are configured in your opencode.json:

{
  "model": {
    "providerID": "anthropic",
    "modelID": "claude-3-5-sonnet-20241022"
  }
}

The plugin respects your configuration and uses the session's default model for all agent executions. This means:

• ✅ You control which models to use via OpenCode config
• ✅ Plugin works with any provider (Anthropic, OpenAI, etc.)
• ✅ No hardcoded model IDs in the plugin
• ✅ Consistent with OpenCode's architecture

Agent Tiers

Agents are organized by complexity tier for potential future enhancements:

| Tier | Agents | Use Case | |------|--------|----------| | LOW | executor-low, explore | Quick lookups, simple tasks | | MEDIUM | executor, researcher, designer | Standard development work | | HIGH | architect, planner, executor-high | Complex design and planning |

Currently, all agents use your configured default model. Tier-based model routing can be enabled in the agent executor if needed.

Configuration

Configure via opencode.json:

{
  "pilot": {
    "routing": {
      "enabled": true,
      "defaultTier": "MEDIUM"
    },
    "features": {
      "astTools": true,
      "lspTools": true,
      "python": true
    }
  }
}

Session Persistence

Session state is automatically saved to .omp/:

• Session metrics and analytics
• Token usage and costs
• Agent and tool usage statistics

License

MIT

Contributing

Contributions welcome! Please read AGENTS.md and SKILLS.md for extending agents and skills.