OpenCode AI Coding Agent Skill

This repository contains the OpenCode AI coding agent skill for the OpenClaw framework.

Overview

OpenCode is an advanced AI coding agent that can:

• Generate code from natural language descriptions
• Review code for bugs, performance issues, and best practices
• Refactor existing code for improved maintainability
• Assist with debugging and fixing issues
• Generate documentation for codebases
• Integrate with ACP (Automated Control Protocol) for web application interaction

Features

• Background Processing: Long-running OpenCode tasks run in the background
• Model Flexibility: Support for various AI models including DeepSeek variants
• Project Context Awareness: Works within specific project directories
• ACO Protocol Support: Automated Control Protocol integration for web apps
• Process Management: Monitor, control, and interact with running OpenCode sessions

Requirements

• OpenCode CLI tool installed and available in PATH
• Appropriate AI model access credentials (if required)

Usage Examples

Basic Code Generation

bash workdir:~/my-project background:true command:"opencode run \"Create a React component for user authentication\""

Using Specific Models

bash workdir:~/my-project background:true command:"opencode -m deepseek/deepseek-coder run \"Implement the API endpoints for user management\""

Code Review

bash workdir:~/my-project background:true command:"opencode run \"Review the changes in src/services and suggest improvements\""

With Helper Script

# Run general OpenCode task
./run.sh run ./my-project "Create a caching layer for the database queries"

# Perform code review
./run.sh review ./my-project "src/utils"

# Fix specific bugs
./run.sh fix ./my-project "memory leaks in data processing module"

Integration Points

This skill integrates with:

• The coding-agent framework
• ACP (Automated Control Protocol) for web application automation
• Background process management for long-running tasks
• Various AI models including DeepSeek, GPT, and others

Safety Guidelines

• Always review AI-generated code before merging to production
• Be specific with prompts to get better results
• Break complex tasks into smaller, focused prompts
• Monitor background processes and terminate if necessary