oh-my-opencode - SKILL.md Agent Skill

name: oh-my-opencode description: Use when working with OpenCode AI coding agent and oh-my-opencode harness to install, configure, and leverage its advanced features including Sisyphus, Hephaestus, Oracle, Librarian, and Explore agents domain: integrations tags:

ai-agent
api
integrations
opencode
third-party

Oh My OpenCode Integration

Overview

Oh My OpenCode is the ultimate harness for OpenCode, providing advanced multi-model orchestration, curated agents, background task execution, and comprehensive tooling. This skill enables OpenClaw to leverage OpenCode's full potential with all the power of oh-my-opencode's agent ecosystem.

When to Use

Installing or configuring OpenCode with oh-my-opencode
Operating OpenCode from OpenClaw environment
Leveraging advanced agents (Sisyphus, Hephaestus, Oracle, Librarian, Explore)
Running background agents and parallel tasks
Using LSP/AST tools, MCP servers, and Claude Code compatibility
Implementing autonomous deep work with goal-oriented execution

Quick Start

Install OpenCode and oh-my-opencode:

Use the oh-my-opencode-installation skill for detailed setup
Or paste this to your LLM agent:

Install and configure oh-my-opencode:
https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/refs/heads/master/docs/guide/installation.md

Initialize for your project:
```
cd /path/to/project
opencode
/init
```
Load this skill in OpenClaw:
- Load the oh-my-opencode skill
- Configure according to your needs

Core Agents

Primary agent handles core task execution
Validator agent checks output quality
Reporter agent formats and delivers results
Each agent operates with clear input/output contracts

Sisyphus — Main Orchestrator

The primary agent that coordinates all others
Uses Prometheus (planner) and Metis (plan consultant)
Executes with "ultrawork" mode for autonomous completion
Delegates to specialized agents based on task requirements

Hephaestus — Autonomous Deep Worker

Goal-oriented execution (GPT 5.3 Codex Medium)
Explores thoroughly before acting (2-5 parallel agents)
End-to-end completion with verification
Pattern matching for codebase consistency

Oracle — Architecture & Debugging

High-IQ strategic consultation (GPT 5.2)
Complex debugging after 2+ failed attempts
Architecture design decisions
Multi-system tradeoffs analysis

Librarian — Documentation & Search

Official docs and open source implementations (GLM-4.7)
Real-time source code digestion
External library patterns and best practices

Explore — Fast Codebase Exploration

Contextual grep for rapid codebase analysis
Pattern discovery across multiple modules
Quick file and symbol location

Installation Methods

This section covers installation methods for the oh-my-opencode skill. Key operations include input validation, core processing, and output verification. Refer to the skill overview for detailed usage instructions.

For Humans

Install and configure oh-my-opencode:
https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/refs/heads/master/docs/guide/installation.md

For LLM Agents

curl -s https://raw.githubusercontent.com/code-yeongyu/oh-my-opencode/refs/heads/master/docs/guide/installation.md

Manual Installation

# Install OpenCode
curl -fsSL https://opencode.ai/install | bash

# Install oh-my-opencode
npm install -g oh-my-opencode

# Or use npm directly
npm install -g @opencode-ai/plugin

Configuration

{
  "enabled": true,
  "autoRun": false,
  "timeout": 30000,
  "retries": 3
}

Set environment variables as needed for authentication and endpoints.

Config Locations

Project: .opencode/oh-my-opencode.jsonc or .opencode/oh-my-opencode.json
User: ~/.config/opencode/oh-my-opencode.jsonc or ~/.config/opencode/oh-my-opencode.json

Key Configuration Options

Agents: Override models, temperatures, prompts for each agent
Built-in Skills: playwright, git-master, and more
Background Tasks: Concurrency limits per provider/model
Categories: Domain-specific task delegation
Hooks: 25+ configurable hooks
MCPs: Built-in websearch (Exa), context7, grep_app
LSP: Full LSP support with refactoring tools

Features

Core operation execution with comprehensive error handling
Input validation and output quality assurance
Integration with existing workflows and toolchains
Detailed logging for debugging and audit trails

Agent Orchestration

Multi-model support with intelligent routing
Domain-specific task categories (visual, business-logic, etc.)
Fallback chains and model recommendations

Background Execution

Run multiple agents in parallel like a real dev team
Async task completion
Todo continuation enforcement

Tools & Refactoring

Full LSP / AST Tools (refactoring, rename, diagnostics)
Hash-anchored edit tool (LINE#ID format)
AST-aware code search and replacement

Claude Code Compatibility

Commands, Agents, Skills, MCP, Hooks
PreToolUse, PostToolUse, UserPromptSubmit, Stop hooks
Full compatibility layer

Session Management

List, read, search, analyze session history
Session continuity across agents
Session-based task continuation

Workflow Integration

Receive input and validate format
Route to appropriate handler based on input type
Execute core operation with monitoring
Transform output to expected format
Return results or trigger follow-up actions

With OpenClaw

Load oh-my-opencode skill in OpenClaw
Ensure OpenCode is installed and configured
Use agents as needed for tasks
Leverage background execution for parallel work

Best Practices

Start with Sisyphus for complex tasks
Use Hephaestus for autonomous deep work
Consult Oracle for architecture decisions
Use Librarian for unfamiliar libraries
Employ Explore for rapid codebase analysis

Common Patterns

Parallel exploration: Fire 2-5 agents simultaneously
Deep work: Use Hephaestus for thorough research first
Debugging: Oracle after 2+ failed attempts
Documentation: Librarian for official docs and OSS patterns

Troubleshooting

Symptom	Cause	Fix
Operation times out	Network or service issue	Check connectivity and retry
Permission denied	Missing credentials	Verify API keys and access tokens
Invalid output	Input format mismatch	Validate input against expected schema

Plugin Not Loading

Verify config file location and format
Check permissions and dependencies
Review error logs: opencode logs

Agent Issues

Check model availability and API keys
Review agent-specific configuration
Validate permissions settings

Performance Issues

Reduce concurrency limits
Disable unused hooks
Optimize model selections

Safety & Considerations

This section covers safety & considerations for the oh-my-opencode skill. Key operations include input validation, core processing, and output verification. Refer to the skill overview for detailed usage instructions.

Terms of Service

OpenCode works with Claude, ChatGPT, and Gemini subscriptions
Some plugins may have ToS implications
Review official documentation for latest guidance

Security

Use official installation sources only
Verify plugin origins before installing
Review permissions before loading skills

The Process

Verify prerequisites – Check Node.js 18+, OpenCode and oh-my-opencode are installed
Install or verify installation – Use oh-my-opencode-installation skill for guided setup
Configure agents – Set models, temperatures, and permissions in config location
Choose the right agent – Sisyphus for coordination, Hephaestus for deep work, Oracle for architecture, Librarian for docs, Explore for exploration
Load skill in OpenClaw – Ensure the skill is accessible to your agent environment
Execute tasks – Use agents with appropriate task categories for domain-specific work
Monitor and iterate – Watch agent output, adjust configuration as needed

When NOT to Use

Task is outside your authorization scope
You need to implement controls (use implementing-* skills)
Task is about analysis, not action (use analyzing-* skills)
You don't have access to target systems
Task requires compliance expertise (consult professionals)
Task is about defense, not offense (use defensive skills)

Red Flags

Trying to use without OpenCode – Oh My OpenCode is a harness on top of OpenCode, not a standalone tool
Using agents incorrectly – Assigning simple tasks to Sisyphus (overkill) or complex tasks to quick category (under-scope)
Ignoring fallback chains – Relying on single model when fallbacks configured for reliability
Skipping verification – Not running lsp_diagnostics, build, or tests after agent completes
Confusing agent capabilities – Using Librarian for code writing (use Hephaestus) or Explore for deep analysis (use Librarian)
Overriding built-in skills – Removing playwright, git-master, which are critical for OpenCode workflows

Verification

Installation verified: Run opencode --version and oh-my-opencode --help - both should execute without error
Agent configuration: Check opencode config get agents shows configured agents with models
Skill loaded: Invoke skill tool and verify content loads without errors
Agent selection: /agent sisyphus or /agent hephaestus should switch without error
Feature working: /help shows expected commands, /status shows valid session state
Agent output: Agent responds with domain-appropriate guidance, not generic responses
Background tasks: Parallel agent launches complete within timeout, results accessible

References

Official Documentation: https://opencode.ai/docs/
Oh My OpenCode: https://github.com/code-yeongyu/oh-my-opencode
Installation Guide: https://github.com/code-yeongyu/oh-my-opencode/blob/master/docs/guide/installation.md
Features: https://github.com/code-yeongyu/oh-my-opencode/blob/master/docs/features.md
Configuration: https://github.com/code-yeongyu/oh-my-opencode/blob/master/docs/configurations.md
Agent-Model Matching: https://github.com/code-yeongyu/oh-my-opencode/blob/master/docs/guide/agent-model-matching.md

Related Skills

oh-my-opencode-installation: Detailed installation guide
oh-my-opencode-agents: Deep dive into each agent
oh-my-opencode-configuration: Configuration options
oh-my-opencode-features: Complete features list
opencode-cli: OpenCode CLI usage

Process

Analyze the task requirements
Apply domain expertise
Verify output quality