dotopencode

OpenCode agent system configuration for AI-assisted software development workflows.

Overview

This repository contains a comprehensive OpenCode configuration implementing a multi-agent system for software development, testing, documentation, and DevOps workflows. It provides structured, reproducible AI-assisted development through agents, skills, and commands.

Structure

.
├── AGENTS.md                  # Core agent system documentation
├── agents/                    # Agent definitions (17 agents)
├── commands/                  # Slash commands (47 commands)
├── skills/                    # Skill library (154 skills)
├── plugins/                   # OpenCode plugins and extensions
├── scripts/                   # Automation scripts
├── tests/                     # Test suite
├── opencode.json             # Core configuration
└── oh-my-opencode.jsonc      # Extended configuration

Components

Agents (17)

Specialized agents for domain-specific tasks:

• Senior-Engineer - Software engineering and implementation
• QA-Engineer - Testing strategy and quality assurance
• Tech-Lead - Architecture and technical leadership
• DevOps - CI/CD and infrastructure
• Security-Engineer - Security auditing
• Data-Analyst - Data analysis and metrics
• Writer - Documentation and content
• Editor - Editorial review
• Knowledge Base Curator - Knowledge management
• VHS-Director - Terminal recordings
• Embedded-Engineer - Embedded systems
• Researcher - Research and investigation
• Code-Reviewer - PR review
• SysOp - System operations
• Linux-Expert - Linux administration
• Nix-Expert - Nix package management
• Model-Evaluator - Model testing

Commands (47)

Workflow automation commands including:

• /commit - AI-assisted git commits
• /pr - Pull request creation
• /test - Test generation and execution
• /review - Code review
• /refactor - Refactoring workflows
• /debug - Debugging assistance
• /research - Investigation workflows
• /vhs - Terminal recording
• And 39 more...

Skills (154)

Comprehensive skill library covering:

• Development: golang, ruby, javascript, cpp, python
• Testing: ginkgo-gomega, jest, cypress, rspec, cucumber
• Architecture: design patterns, clean code, api-design
• DevOps: docker, aws, heroku, nix, automation
• Documentation: writing-style, api-documentation, accessibility-writing
• Workflows: bdd-workflow, git-master, pr-review-workflow
• Communication: email-communication, incident-communication
• And 140+ more...

Installation

Prerequisites

• OpenCode installed
• Git
• Node.js/Bun (for plugins)

Setup

# Clone the repository
git clone https://github.com/baphled/dotopencode.git

# Symlink to OpenCode config directory
ln -s "$(pwd)/dotopencode" ~/.config/opencode

# Or add as git submodule to your dotfiles
cd ~/dotfiles
git submodule add https://github.com/baphled/dotopencode.git .config/opencode

Usage

Basic Workflow

# Use slash commands
/commit "feat: add new feature"
/test path/to/file.go
/pr "Feature implementation"

# Or invoke agents directly via OpenCode CLI
opencode --agent Senior-Engineer "Implement authentication"

Agent Delegation

Agents automatically delegate to specialists:

// Orchestrator delegates to specialist
task(
  subagent_type="Senior-Engineer",
  load_skills=["golang", "clean-code", "bdd-workflow"],
  description="Implement user service",
  prompt="Create user service with CRUD operations following repository pattern"
)

Skill Loading

Skills use progressive disclosure - load only what's needed:

// Mandatory always-active skills
["pre-action", "memory-keeper", "discipline"]

// Task-specific skills
task(
  category="quick",
  load_skills=["golang", "ginkgo-gomega"],
  description="Add unit tests"
)

Configuration

opencode.json

Core OpenCode configuration with MCP server setup:

{
  "mcpServers": {
    "memory": {...},
    "vault-rag": {...},
    "websearch": {...}
  }
}

oh-my-opencode.jsonc

Extended configuration with:

• Model routing (T1/T2/T3 tiers)
• Agent definitions
• Provider health monitoring
• Category-based delegation

Key Features

Three Pillar Model

1. Always-Active Skills - Mandatory skills loaded on every task
2. Parallel Execution - Independent tasks run simultaneously
3. Progressive Disclosure - Load skills only when needed

Commit Workflow

Hybrid approach combining git-master planning + make ai-commit execution:

# Plan commit structure
/commit
# Executes: task(load_skills=["git-master"], ...)
# Writes to tmp/commit.txt
# Runs: make ai-commit FILE=tmp/commit.txt

Model Routing

Three-tier system optimized for task complexity:

• T1 (gpt-5-mini, Haiku) - Exploration, search
• T2 (gpt-4.1, Sonnet 4) - Implementation, tests
• T3 (gpt-5.2, Opus 4.6) - Architecture, novel problems

Development

Running Tests

# BATS tests for skill import pipeline
bun test

# Plugin tests
cd plugins && bun test

Adding Skills

/new-skill "skill-name" "Description of skill"

Adding Agents

Create agent definition in agents/:

---
name: agent-name
model: claude-sonnet-4-5
tier: T2
---

## Description
...

Contributing

See CONTRIBUTING.md for contribution guidelines.

License

MIT License - see LICENSE

Acknowledgments

Built with OpenCode and inspired by:

• The Cascade methodology
• AI-assisted development workflows
• Knowledge management best practices

---

Extracted from: baphled/dotfiles
Created by: Yomi Colledge (@baphled)