OpenCode Baseline
A production-ready template for OpenCode AI assistant with pre-configured agents, skills, commands, hooks, and context.
Table of Contents
Installation
Prerequisites
- Git - Required for cloning and updates
- Node.js 18+ - Required for hooks (TypeScript-based)
- OpenCode - The AI assistant (opencode.ai)
Quick Install (One-Line)
curl -fsSL https://raw.githubusercontent.com/48Nauts-Operator/opencode-baseline/main/install.sh | bash
Install Options
| Option |
Description |
--global |
Install to ~/.config/opencode/ (global for all projects) |
--init |
Create project-specific files only (OPENCODE.md, AGENTS.md) |
--project |
Install .opencode/ to current project (legacy default) |
--no-env |
Skip copying .env.example |
--no-opencode-md |
Skip copying OPENCODE.md template |
--force, -f |
Overwrite existing .opencode directory |
--branch, -b |
Use specific branch (default: main) |
# Examples
curl -fsSL <url>/install.sh | bash -s -- --global # Global install
curl -fsSL <url>/install.sh | bash -s -- --init # Project init only
curl -fsSL <url>/install.sh | bash -s -- --force # Force overwrite
curl -fsSL <url>/install.sh | bash -s -- --branch develop --no-env
Global vs Project Install
| Mode |
Location |
Use Case |
--global |
~/.config/opencode/ |
Skills available in ALL projects |
--project |
./.opencode/ |
Project-specific customization |
--init |
./OPENCODE.md, ./AGENTS.md |
Add context files to existing project |
Manual Install
# Clone the repository
git clone https://github.com/48Nauts-Operator/opencode-baseline.git
# Copy to your project
cp -r opencode-baseline/.opencode /path/to/your/project/
cp opencode-baseline/OPENCODE.md /path/to/your/project/
cp opencode-baseline/.env.example /path/to/your/project/.env
Post-Install Setup
Configure Project Context
# Edit the project context file with your project details
nano .opencode/context/project/project-context.md
Set Up Environment Variables
cp .env.example .env
nano .env
Customize OPENCODE.md
# Edit project description for AI context
nano OPENCODE.md
Environment Variables
Create a .env file with your API keys:
# AI Provider Keys (at least one required)
OPENAI_API_KEY=sk-...
ANTHROPIC_API_KEY=sk-ant-...
GEMINI_API_KEY=AIza...
# Optional: GitHub integration
GITHUB_TOKEN=ghp_...
Updating
Update to the latest version while preserving your customizations:
./update.sh
| Option |
Description |
--branch, -b |
Use specific branch |
--preserve, -p |
Additional path to preserve (repeatable) |
--full |
Full update, don't preserve anything |
Default preserved paths:
context/project/ - Your project-specific contextopencode.json - Your configurationOPENCODE.md - Your project description
Uninstalling
./uninstall.sh
| Option |
Description |
--force, -f |
Skip confirmation prompt |
--keep-context, -k |
Keep .opencode/context/project/ |
Verifying Installation
# Check directory structure
ls -la .opencode/
# Verify version
cat .opencode/VERSION
# Check configuration
cat .opencode/opencode.json | head -20
What's Included
| Component |
Count |
Description |
| Skills |
55 |
Reusable workflows + domain knowledge (K8s, Security, LLM, MLOps, Research, etc.) |
| Agents |
35 |
Specialized AI personas (language experts, DevOps, architecture, etc.) |
| Commands |
18 |
Slash commands for operations + incident response + orchestration |
| Hooks |
6 |
Lifecycle hooks (pre/post tool, session, context monitor, pre-compact, prompt) |
| Plugins |
3 |
Extensions (agent validator, notifications) |
| Tools |
4 |
Custom tools (Gemini image editing, env management) |
| Context Files |
20+ |
Domain knowledge and standards |
v0.6.0 Update: Added 7 amp-skills from snarktank/amp-skills, Ralph autonomous loop system, global install support, and 12 Claude Code commands
v0.5.0 Update: Consolidated all hooks to TypeScript/npm package with 6 CLI commands: context monitoring, pre-compact backup, user prompt logging. No more Python dependency!
v0.4.0 Update: Added 10 research skills from GhostScientist/skills - paper implementation, research workflows, and academic tools
v0.3.0 Update: Added 27 new skills and 7 new agents from wshobson/agents (23k ⭐)
Skills
Reusable workflows that teach AI how to perform specific tasks. Located in .opencode/skill/.
| Skill |
Description |
Trigger Phrases |
| prd |
Generate Product Requirements Documents |
"create a prd", "plan this feature", "requirements for" |
| ralph |
Convert PRDs to JSON for autonomous execution |
"convert this prd", "ralph json", "turn this into ralph format" |
| dev-browser |
Browser automation with persistent page state |
"go to [url]", "click on", "screenshot", "scrape" |
| compound-engineering |
Plan -> Work -> Review -> Compound development loop |
"plan this feature", "systematic development" |
| docx |
Word document creation, editing, tracked changes |
Work with .docx files |
| pdf |
PDF manipulation: extract, merge, split, create |
Work with PDF files |
| frontend-design |
Create distinctive UI/UX interfaces (anti-AI-slop) |
"build web component", "create page" |
Core Skills
| Skill |
Description |
Workflow |
| tdd |
Test-Driven Development with RED-GREEN-REFACTOR cycle |
Development |
| code-review |
Systematic code review checklist for quality and security |
Review |
| git-release |
Create consistent releases with changelogs and version bumping |
GitHub |
| systematic-debugging |
4-phase debugging methodology for finding root causes |
Debugging |
| skill-creator |
Meta-skill for creating new OpenCode skills |
Meta |
| changelog-generator |
Generate user-friendly changelogs from git commits |
Release |
| mcp-builder |
Build MCP servers for AI agent integrations |
Development |
| file-organizer |
Intelligently organize files with duplicate detection |
Productivity |
| pr-create |
Create well-structured pull requests with proper descriptions |
Git |
Kubernetes Skills (skill/kubernetes/)
| Skill |
Description |
| k8s-manifest-generator |
Production-ready K8s manifests (Deployments, Services, ConfigMaps) |
| helm-chart-scaffolding |
Helm chart creation, templating, multi-environment deployment |
| gitops-workflow |
ArgoCD/Flux GitOps automation, declarative deployments |
| k8s-security-policies |
Network policies, RBAC patterns, security hardening |
Security Skills (skill/security/)
| Skill |
Description |
| sast-configuration |
SAST tool setup (Semgrep, SonarQube, CodeQL), custom rules |
| attack-tree-construction |
Threat modeling, attack path visualization |
| stride-analysis-patterns |
STRIDE threat modeling methodology |
| security-requirement-extraction |
Security requirements from specifications |
| threat-mitigation-mapping |
Mapping threats to mitigations |
LLM Development Skills (skill/llm-dev/)
| Skill |
Description |
| langchain-architecture |
LangChain agents, chains, memory, tool integration |
| rag-implementation |
Retrieval-Augmented Generation patterns |
| prompt-engineering-patterns |
Prompt optimization, few-shot learning, chain-of-thought |
| embedding-strategies |
Vector embeddings for semantic search |
| vector-index-tuning |
Vector database optimization |
| similarity-search-patterns |
Semantic similarity algorithms |
| hybrid-search-implementation |
Combining keyword + semantic search |
| llm-evaluation |
LLM output quality assessment |
Developer Essentials (skill/developer-essentials/)
| Skill |
Description |
| git-advanced-workflows |
Rebase, cherry-pick, bisect, worktrees, reflog |
| auth-implementation-patterns |
JWT, OAuth2, session management, security |
| error-handling-patterns |
Error handling across languages |
| e2e-testing-patterns |
End-to-end testing strategies |
| sql-optimization-patterns |
Query optimization, indexing |
| monorepo-management |
Monorepo tooling and workflows |
| bazel-build-optimization |
Bazel build system patterns |
| turborepo-caching |
Turborepo caching strategies |
| nx-workspace-patterns |
Nx monorepo patterns |
MLOps Skills (skill/mlops/)
| Skill |
Description |
| ml-pipeline-workflow |
End-to-end MLOps: data prep → training → deployment → monitoring |
| Skill |
Description |
| implement-paper-from-scratch |
Step-by-step guide to implementing research papers from scratch |
| paper-to-intuition |
Transform academic papers into deep, multi-layered understanding |
| research-question-refiner |
Refine vague research interests into concrete, tractable questions |
| research-taste-developer |
Develop intuition for what makes research "good" vs "incremental" |
| reviewer-2-simulator |
Critique paper drafts as a skeptical reviewer would |
| experiment-design-checklist |
Generate rigorous experiment designs with controls and metrics |
| turn-this-feature-into-a-blog-post |
Generate technical blog posts from code implementations |
| ios-app-icon-generator |
Generate complete iOS app icon sets with all required sizes |
| create-watchos-version |
Plan watchOS companion apps from existing iOS/macOS projects |
| hugging-face-space-deployer |
Deploy Hugging Face Spaces for ML models (Gradio, Streamlit) |
Using Skills
Skills are invoked automatically based on context or explicitly:
# Invoke TDD workflow
"Let's implement this feature using TDD"
# Create a skill
"Use skill-creator to build a deployment skill"
# Create a PRD
/prd Create a user authentication system
# Convert PRD to Ralph JSON
/ralph Convert my PRD to JSON format
Ralph Autonomous Loop
Ralph is an autonomous AI agent loop that executes user stories from a PRD one at a time. Located in ralph/.
How It Works
- Create a PRD using the
/prd skill
- Convert to JSON using the
/ralph skill (creates prd.json)
- Run
ralph.sh to autonomously execute each user story
Usage
# Navigate to your project
cd your-project
# Run Ralph (auto-detects amp/opencode/claude CLI)
~/.config/opencode/ralph/ralph.sh
# With max iterations limit
~/.config/opencode/ralph/ralph.sh 20
Ralph Files
| File |
Description |
ralph.sh |
Universal adapter script (works with amp, opencode, claude) |
prompt.md |
Agent instructions for each iteration |
prd.json.example |
Example PRD format |
PRD Format
{
"project": "MyApp",
"branchName": "ralph/feature-name",
"description": "Feature description",
"userStories": [
{
"id": "US-001",
"title": "Story title",
"description": "As a user, I want...",
"acceptanceCriteria": ["Criterion 1", "Criterion 2"],
"priority": 1,
"passes": false,
"notes": ""
}
]
}
Key Principles
- One story per iteration: Ralph spawns fresh context each time
- Stories must be small: Completable in one context window
- Dependency order: Schema -> Backend -> UI
- Browser verification: UI stories require dev-browser verification
Agents
Specialized AI personas for different domains. Located in .opencode/agent/.
Core Agents
| Agent |
File |
Description |
| OpenAgent |
core/openagent.md |
Main orchestrator - routes requests to specialists |
| OpenCoder |
core/opencoder.md |
General-purpose coding agent |
Development Agents
| Agent |
File |
Description |
| Backend Specialist |
development/backend-specialist.md |
APIs, databases, server logic |
| Frontend Specialist |
development/frontend-specialist.md |
UI, UX, components, styling |
| DevOps Specialist |
development/devops-specialist.md |
CI/CD, deployment, containerization |
| Codebase Agent |
development/codebase-agent.md |
Codebase understanding and navigation |
| Event Sourcing Architect |
development/event-sourcing-architect.md |
Event sourcing, CQRS, saga orchestration |
| GraphQL Architect |
development/graphql-architect.md |
GraphQL schema design, resolvers |
| TDD Orchestrator |
development/tdd-orchestrator.md |
Test-driven development coordination |
Architecture Agents (NEW)
| Agent |
File |
Description |
| C4 Context |
architecture/c4-context.md |
System context diagrams |
| C4 Container |
architecture/c4-container.md |
Container-level architecture |
| C4 Component |
architecture/c4-component.md |
Component diagrams |
| C4 Code |
architecture/c4-code.md |
Code-level documentation |
Language Specialists
| Agent |
File |
Triggers |
| Python Pro |
languages/python-pro.md |
.py files |
| JavaScript Pro |
languages/javascript-pro.md |
.js, .mjs files |
| TypeScript Pro |
languages/typescript-pro.md |
.ts, .tsx files |
| Rust Pro |
languages/rust-pro.md |
.rs files |
| Go Pro |
languages/golang-pro.md |
.go files |
| SQL Pro |
languages/sql-pro.md |
.sql files |
Infrastructure Agents
| Agent |
File |
Triggers |
| Terraform Specialist |
infrastructure/terraform-specialist.md |
"terraform", "infrastructure", "IaC" |
| Performance Engineer |
infrastructure/performance-engineer.md |
"optimize", "performance", "slow" |
| DevOps Troubleshooter |
infrastructure/devops-troubleshooter.md |
"debug", "error", "not working" |
| Security Specialist |
infrastructure/security-specialist.md |
"security", "vulnerability", "CVE" |
Content Agents
| Agent |
File |
Description |
| Copywriter |
content/copywriter.md |
Marketing copy, messaging |
| Technical Writer |
content/technical-writer.md |
Documentation, guides |
Data Agents
| Agent |
File |
Description |
| Data Analyst |
data/data-analyst.md |
Data analysis and visualization |
Subagents (Delegated Tasks)
| Agent |
File |
Description |
| Build Agent |
subagents/code/build-agent.md |
Build verification and validation |
| Coder Agent |
subagents/code/coder-agent.md |
Code execution tasks |
| Reviewer |
subagents/code/reviewer.md |
Code review and quality checks |
| Tester |
subagents/code/tester.md |
Test authoring and TDD |
| Codebase Pattern Analyst |
subagents/code/codebase-pattern-analyst.md |
Pattern discovery in codebases |
| Task Manager |
subagents/core/task-manager.md |
Task breakdown and tracking |
| Documentation |
subagents/core/documentation.md |
Documentation generation |
| Context Retriever |
subagents/core/context-retriever.md |
Context file search and retrieval |
System Builder Subagents
| Agent |
File |
Description |
| Agent Generator |
subagents/system-builder/agent-generator.md |
Generate new agent definitions |
| Command Creator |
subagents/system-builder/command-creator.md |
Create custom slash commands |
| Context Organizer |
subagents/system-builder/context-organizer.md |
Organize context files |
| Domain Analyzer |
subagents/system-builder/domain-analyzer.md |
Analyze project domains |
| Workflow Designer |
subagents/system-builder/workflow-designer.md |
Design workflow definitions |
Utility Subagents
| Agent |
File |
Description |
| Image Specialist |
subagents/utils/image-specialist.md |
Image editing and analysis |
Meta Agents
| Agent |
File |
Description |
| System Builder |
meta/system-builder.md |
Build complete context systems |
Commands
Slash commands for common operations. Located in .opencode/command/.
Core Commands
| Command |
Description |
/clean |
Format, lint, and clean up code (Prettier, ESLint, import sorting) |
/commit |
Create conventional commits with emoji |
/git-flow |
Start new development context (branch/worktree) |
/git-safety |
Run safety checks before merging (secrets, debug, conflicts) |
/git-commit |
Smart conventional commit generation with auto-detection |
/optimize |
Analyze and optimize code for performance and security |
/test |
Run the complete testing pipeline |
/worktrees |
Manage git worktrees for parallel development |
/context |
Analyze and understand project context and structure |
/build-context-system |
Create complete AI context architecture |
Incident Response Commands (NEW)
| Command |
Description |
/smart-fix |
4-phase AI-powered debugging: analysis → root cause → fix → verification |
/incident-response |
Production incident triage and resolution workflow |
Agent Orchestration Commands (NEW)
| Command |
Description |
/multi-agent-optimize |
Multi-agent performance optimization across system layers |
/improve-agent |
Agent improvement and optimization workflow |
Architecture Commands (NEW)
| Command |
Description |
/c4-architecture |
Generate C4 model documentation from code |
/full-stack-feature |
End-to-end feature: backend → frontend → tests → security → deploy |
Prompt Engineering Commands
| Command |
Description |
/prompt-enhancer |
Enhance prompts for better AI responses |
/prompt-optimizer |
Optimize prompts for specific models |
Usage Examples
# Start a feature branch
/git-flow --medium "Add user authentication"
# Check before merging
/git-safety --strict
# Smart commit
/git-commit
# Clean and format
/clean
Claude Code Commands (NEW)
Additional commands for Claude Code users. Located in global/claude-code/commands/.
| Command |
Description |
/five-whys |
Root cause analysis using Five Whys methodology |
/design-review |
Comprehensive design review of git changes |
/lyra |
AI prompt optimization specialist |
/utrathink |
Multi-agent orchestration (4 sub-agents) |
/analyze-issue |
GitHub issue to implementation spec |
/refactor |
Code refactoring with best practices |
/commit-fast |
Conventional commits with smart staging |
/tdd |
Test-Driven Development enforcement |
/prompt_sec |
Security audit for APIs |
/reload |
Reload project memory from CLAUDE.md |
/analyse_codebase |
Full codebase analysis and documentation |
/clean |
Code cleanup, formatting, linting |
Installation for Claude Code
# Copy commands to Claude Code config
cp -r global/claude-code/commands/* ~/.claude/commands/
Hooks
Lifecycle hooks that run at specific points during AI sessions. All hooks are TypeScript-based and bundled in the npm package.
| Hook |
When |
Can Block |
Purpose |
opencode-pre-tool |
Before each tool call |
Yes (exit 2) |
Security validation, command blocking |
opencode-post-tool |
After each tool call |
No |
Logging, metrics, error tracking |
opencode-session-start |
Session begins |
No |
Load context, git status, initialization |
opencode-context-monitor |
On notifications |
No |
Track token usage, backup at thresholds |
opencode-pre-compact |
Before compaction |
No |
Backup transcript and todos |
opencode-user-prompt |
On user prompt |
Yes (exit 2) |
Log prompts, optional validation |
Hook Features
Security (opencode-pre-tool):
- Blocks dangerous commands (
rm -rf /, sudo, etc.)
- Prevents access to sensitive files (
.env, credentials)
- Detects secrets in write operations (API keys, tokens)
- Blocks dangerous git operations (force push to main)
Logging (opencode-post-tool):
- Logs all tool usage to JSON files
- Tracks errors and daily statistics
- Monitors session activity
Context Management (opencode-context-monitor):
- Tracks cumulative token usage
- Creates backups at 70% (warning) and 85% (critical)
- Saves pending todos and recent prompts for recovery
Pre-Compact Backup (opencode-pre-compact):
- Backs up transcript before context compaction
- Saves todos state for session recovery
Installation & Configuration
IMPORTANT: OpenCode and Claude Code use DIFFERENT hook systems!
For OpenCode (Recommended)
OpenCode uses a plugin system. The hooks are bundled as a TypeScript plugin in .opencode/plugin/hooks.ts.
Option 1: Use the bundled plugin (default)
The plugin is pre-configured in .opencode/opencode.json:
{
"plugin": [
"hooks"
]
}
This automatically loads .opencode/plugin/hooks.ts which provides all security, logging, and notification features.
Option 2: Install globally and add to user config
npm install -g opencode-baseline-hooks
Then add to ~/.config/opencode/opencode.json:
{
"plugin": [
"opencode-baseline-hooks"
]
}
For Claude Code
Claude Code uses a hooks config with shell commands. Install the npm package and configure in ~/.claude/settings.json:
npm install -g opencode-baseline-hooks
{
"hooks": {
"PreToolUse": [{
"matcher": "",
"hooks": [{
"type": "command",
"command": "opencode-pre-tool"
}]
}],
"PostToolUse": [{
"matcher": "",
"hooks": [{
"type": "command",
"command": "opencode-post-tool"
}]
}],
"SessionStart": [{
"matcher": "",
"hooks": [{
"type": "command",
"command": "opencode-session-start"
}]
}]
}
}
Environment Variables
KOKORO_URL=http://localhost:8880 # Kokoro TTS endpoint
KOKORO_VOICE=bf_emma # Voice selection
OPENCODE_VOICE=off # Disable voice entirely
PROJECT_DIR=/path/to/project # Project root (auto-detected)
Platform Comparison
| Feature |
OpenCode |
Claude Code |
| Config file |
opencode.json |
~/.claude/settings.json |
| Hook system |
"plugin": ["hooks"] |
"hooks": { "PreToolUse": [...] } |
| Hook format |
TypeScript modules |
Shell commands |
| Installation |
Bundled or npm plugin |
npm + config |
Package: npmjs.com/package/opencode-baseline-hooks
Plugins
Extensions that add functionality to OpenCode. Located in .opencode/plugin/.
Agent Validator
Real-time validation of agent behavior against defined rules.
| Tool |
Description |
validate_session |
Validate current session against rules |
check_approval_gates |
Check approval gate enforcement |
export_validation_report |
Export validation report to markdown |
analyze_delegation |
Analyze delegation decisions (4+ file rule) |
analyze_context_reads |
Show context files read during session |
check_context_compliance |
Verify context read before execution |
analyze_agent_usage |
Show agent activity and tool usage |
debug_validator |
Debug validator tracking state |
Additional Plugin Features
- Telegram Notify (
telegram-notify.ts) - Send notifications via Telegram
- Notify (
notify.ts) - Generic notification system
Tools
Custom tools that extend OpenCode capabilities. Located in .opencode/tool/.
Gemini Image Tools
AI-powered image editing and analysis using Google Gemini.
| Tool |
Description |
/gemini |
Edit images with natural language prompts |
/gemini_multiple_edit |
Advanced image editing |
/gemini_multiple_analyze |
Analyze images without editing |
/gemini_edit |
Auto-detection of pasted images |
Setup
# Add to .env
GEMINI_API_KEY=your_api_key_here
Usage
# Edit an image
/gemini "photo.png" "Add a vintage filter" "output.png"
# Analyze code screenshot
/gemini_multiple_analyze "code.png" "What bugs can you spot?"
Environment Tools
Located in .opencode/tool/env/ - Environment variable management utilities.
Prompts System
Multi-model prompt variants with evaluation framework. Located in .opencode/prompts/.
Available Variants
| Variant |
Model Family |
Status |
Best For |
default |
Claude |
Stable |
Production, Claude models |
gpt |
GPT |
Stable |
GPT-4, GPT-4o |
gemini |
Gemini |
Stable |
Gemini 2.0, Gemini Pro |
grok |
Grok |
Stable |
Grok models (free tier) |
llama |
Llama/OSS |
Stable |
Llama, Qwen, DeepSeek |
Testing Variants
# Test with evaluation framework
cd evals/framework
npm run eval:sdk -- --agent=openagent --prompt-variant=llama --suite=smoke-test
# Switch variant permanently
./scripts/prompts/use-prompt.sh --agent=openagent --variant=llama
Creating Custom Variants
- Copy template:
cp .opencode/prompts/openagent/TEMPLATE.md .opencode/prompts/openagent/custom.md
- Edit metadata (YAML frontmatter)
- Customize prompt content
- Test with eval framework
Context System
Domain knowledge and standards. Located in .opencode/context/.
Core Context
| File |
Description |
core/essential-patterns.md |
Essential coding patterns |
core/standards/code.md |
Code standards |
core/standards/docs.md |
Documentation standards |
core/standards/tests.md |
Testing standards |
core/standards/patterns.md |
Design patterns |
core/standards/analysis.md |
Analysis standards |
core/workflows/delegation.md |
Delegation workflow |
core/workflows/review.md |
Review workflow |
core/workflows/sessions.md |
Session management |
core/workflows/task-breakdown.md |
Task breakdown methodology |
core/workflows/design-iteration.md |
Design iteration process |
core/system/context-guide.md |
Context system guide |
Development Context
| File |
Description |
development/react-patterns.md |
React best practices |
development/api-design.md |
API design patterns |
development/clean-code.md |
Clean code principles |
development/animation-patterns.md |
Animation patterns |
development/design-systems.md |
Design system guidelines |
development/design-assets.md |
Design asset management |
development/ui-styling-standards.md |
UI styling standards |
Content Context
| File |
Description |
content/copywriting-frameworks.md |
Copywriting frameworks |
content/tone-voice.md |
Tone and voice guidelines |
Project Context
| File |
Description |
project/project-context.md |
Project-specific context (customize this!) |
System Builder Templates
| File |
Description |
system-builder-templates/SYSTEM-BUILDER-GUIDE.md |
Guide for building systems |
system-builder-templates/orchestrator-template.md |
Orchestrator agent template |
system-builder-templates/subagent-template.md |
Subagent template |
Configuration
Main configuration in .opencode/opencode.json.
Instructions
Files automatically loaded as context:
{
"instructions": [
"OPENCODE.md",
"CONTRIBUTING.md",
"CLAUDE.md",
".cursorrules",
"README.md"
]
}
Permissions
Three permission levels:
| Level |
Behavior |
Examples |
allow |
Execute without confirmation |
git, npm, python, ls |
deny |
Block completely |
rm -rf /, sudo, kill -9 |
ask |
Require user confirmation |
rm, git push --force |
Pre-configured Safety
Allowed (safe operations):
- Git operations, package managers (npm, yarn, pip, cargo)
- Build tools, linters, formatters
- Docker, file read/write, search tools
Denied (dangerous operations):
rm -rf /, sudo, kill -9- Force pushes to main/master
- Disk formatting commands
- Docker system prune
Ask (require confirmation):
- File deletion (
rm)
- Force push to branches
- Package uninstallation
Directory Structure
.
├── global/ # For global installs
│ ├── opencode/ # -> ~/.config/opencode/
│ │ └── skill/ # Global skills (amp-skills)
│ └── claude-code/ # -> ~/.claude/
│ └── commands/ # Claude Code commands
│
├── ralph/ # Autonomous AI agent loop
│ ├── ralph.sh # Universal adapter script
│ ├── prompt.md # Agent instructions
│ └── prd.json.example # Example PRD format
│
├── project-template/ # For --init flag
│ ├── OPENCODE.md # Project description template
│ ├── AGENTS.md # Learnings template
│ └── .opencode/context/project/
│
├── .opencode/
│ ├── opencode.json # Main configuration
│ ├── VERSION # Version number
│ ├── README.md # Internal documentation
│ │
│ ├── agent/ # AI agent definitions
│ │ ├── core/ # Main agents (openagent, opencoder)
│ │ ├── development/ # Dev specialists
│ │ ├── languages/ # Language specialists
│ │ ├── infrastructure/ # DevOps/infra specialists
│ │ ├── content/ # Content creators
│ │ ├── data/ # Data specialists
│ │ ├── meta/ # System builder
│ │ └── subagents/ # Delegated task agents
│ │ ├── code/ # Code-related subagents
│ │ ├── core/ # Core subagents
│ │ ├── system-builder/ # System building subagents
│ │ └── utils/ # Utility subagents
│ │
│ ├── command/ # Slash commands
│ │ ├── git-flow.md
│ │ ├── git-safety.md
│ │ ├── git-commit.md
│ │ ├── clean.md
│ │ ├── commit.md
│ │ ├── optimize.md
│ │ ├── test.md
│ │ ├── worktrees.md
│ │ ├── context.md
│ │ ├── build-context-system.md
│ │ └── prompt-engineering/
│ │
│ ├── context/ # Domain knowledge
│ │ ├── core/ # Essential patterns & standards
│ │ ├── development/ # Dev standards
│ │ ├── content/ # Content standards
│ │ ├── project/ # Project-specific (customize!)
│ │ └── system-builder-templates/
│ │
│ ├── hooks/ # Lifecycle hooks
│ │ ├── pre_tool_use.py
│ │ ├── post_tool_use.py
│ │ └── session_start.py
│ │
│ ├── skill/ # Reusable workflows
│ │ ├── tdd/
│ │ ├── code-review/
│ │ ├── git-release/
│ │ ├── systematic-debugging/
│ │ ├── skill-creator/
│ │ ├── changelog-generator/
│ │ ├── mcp-builder/
│ │ ├── file-organizer/
│ │ └── pr-create/
│ │
│ ├── plugin/ # Extensions
│ │ ├── agent-validator.ts
│ │ ├── telegram-notify.ts
│ │ └── notify.ts
│ │
│ ├── tool/ # Custom tools
│ │ ├── gemini/
│ │ ├── env/
│ │ └── template/
│ │
│ ├── prompts/ # Model-specific variants
│ │ ├── core/
│ │ ├── content/
│ │ ├── data/
│ │ └── development/
│ │
│ ├── ai-docs/ # AI documentation
│ │ ├── anthropic_*.md
│ │ ├── openai_*.md
│ │ └── cc_hooks_docs.md
│ │
│ └── node_modules/ # Dependencies
│
├── OPENCODE.md # Project context template
├── .env.example # Environment template
├── .gitignore
├── install.sh # One-line installer
├── update.sh # Update script
└── uninstall.sh # Uninstall script
Ecosystem: Plugins, Tools & Extensions
A curated collection of OpenCode ecosystem tools discovered via parallel librarian agents.
Core Plugins (Must-Have)
| Plugin |
Stars |
Description |
Install |
| oh-my-opencode |
4,135 |
Async subagents, LSP/AST tools, curated agents (Sisyphus, Oracle, Librarian) |
npm i -g oh-my-opencode |
| opencode-roadmap |
16 |
Persistent todo lists across sessions, multi-agent coordination |
npm i -g opencode-roadmap |
| opencode-tokenscope |
31 |
Token usage analysis & cost tracking per model |
npm i -g opencode-tokenscope |
| opencode-dynamic-context-pruning |
233 |
Intelligently prunes context to save tokens |
npm i -g opencode-dynamic-context-pruning |
| opencode-sessions |
71 |
Multi-agent session management & collaboration |
npm i -g opencode-sessions |
Multi-Agent Orchestration
| Plugin |
Stars |
Description |
| claude-squad |
5,454 |
Manage multiple AI terminal agents in parallel |
| superset |
334 |
Run dozens of agents in parallel using git worktrees |
| swarm-tools |
250 |
Multi-agent swarm coordination with learning |
| opencode-background |
19 |
Manage and communicate with background tasks |
| kimaki |
118 |
Control OpenCode agents via Discord (voice channels too!) |
Context & Memory
Development Tools
Authentication Providers
Skills & Workflows
Monitoring & Analytics
Code Quality & Review
UI & Notifications
Environment & Config
Agent Collections
Hooks & Automation
Templates & Starters
Utilities
MCP Servers (Swiss Army Knives)
All-in-One Powerhouses
| Server |
Stars |
Description |
| PAL MCP Server |
10,685 |
Multi-model: Claude/Gemini/OpenAI/Grok/Ollama |
| Activepieces |
20,134 |
~400 MCP servers for AI agents |
| MindsDB |
38,138 |
"The only MCP Server you'll ever need" |
| FastMCP |
21,568 |
Fast, Pythonic MCP server builder |
| Context7 |
40,504 |
Up-to-date code documentation for LLMs |
Database Access
| Server |
Stars |
Description |
| DBHub |
1,809 |
Zero-dep, multi-DB (Postgres, MySQL, SQLite, etc.) |
| MongoDB MCP |
868 |
MongoDB Atlas integration |
| Qdrant MCP |
1,150 |
Vector database |
| Neo4j MCP |
859 |
Graph database |
Browser Automation
API Integrations
Search & Data
| Server |
Stars |
Description |
| Exa MCP |
3,474 |
Web search and crawling |
| Firecrawl MCP |
5,152 |
Web scraping and search |
| Arxiv MCP |
1,985 |
Search and analyze research papers |
Productivity
Curated Lists
Resources
Contributing
- Fork the repository
- Create a feature branch
- Make your changes
- Run tests if applicable
- Submit a pull request
License
MIT