Build configurable plugin system for custom behaviors and efficiency

[darrenhinde]

Overview

Implement a comprehensive plugin architecture following Claude Code plugin patterns to enable custom behaviors, improved efficiency, and self-contained experiences while maintaining configurability.

Goals

• Implement plugin architecture following Claude Code plugin patterns
• Support slash commands, agents, skills, hooks, and MCP servers
• Enable concurrent subagent execution for improved efficiency
• Add plugin manifest system (.opencode-plugin/plugin.json)
• Create plugin loading and lifecycle management
• Build plugin validation and debugging tools
• Support both standalone (.opencode/) and plugin distribution modes
• Enable self-contained plugin experience with access to models, prompts, and configuration

Key Features

Plugin Structure

• .opencode-plugin/plugin.json manifest
• commands/ for slash commands
• agents/ for custom agents
• skills/ for agent skills
• hooks/ for event handlers
• .mcp.json for MCP server configurations

Concurrent Execution

• Support parallel subagent execution for independent tasks
• Implement efficient multi-agent coordination patterns
• Add dependency resolution for sequential tasks

Self-Contained Experience

• Allow plugins to bundle models, prompts, and configurations
• Enable customization of bundled components
• Support plugin-specific settings and overrides
• Provide access to model selection and prompt customization within plugins

Distribution

• npm packages for plugin distribution
• Plugin marketplace/registry support
• Version management and updates

Related Issues

• Add Skills/Abilities system to OpenCode #33, feat: Skills #29 (Skills/Abilities system)

Success Criteria

• Plugin manifest system implemented
• Concurrent subagent execution working
• Plugin loading and lifecycle management complete
• Validation and debugging tools available
• Self-contained plugin mode with customization support
• Documentation and examples provided