Open Source

An Agent System for opencode

66 specialized agents that act on your intent

Start BuildingView on GitHub

66

Agents

62

Project Skills

11

Meta-Skills

5

Primary Agents

What's NewLatest additions to the toolkit
View all changes →

testVerifySettings

Six booleans are now the sole Playwright gate — replaces testIntensity, storyAssessment, rigorProfile, and E2E deferral.

Verification State Reset

Stale verification contracts no longer carry over between tasks — state resets at every task boundary.

OAuth Diagnostic Skill

New skill diagnoses OAuth redirect_uri mismatches across any provider and app type — desktop and web.

Unified State Model

Single activeWork object replaces separate PRD and ad-hoc state — stories, decisions, and resume all in one place.

Agents that act

Traditional AI assistants respond once and stop. An agentic AI is different—it takes action. It reads files, runs commands, browses the web, and iterates until the work is done.

This toolkit includes 66 specialized agents. Some write code. Some review it. Some run tests or deploy infrastructure. Primary agents call on specialists as needed—expert help, orchestrated automatically.

A team of focused collaborators. Your ideas, realized.

Explore agents →

Primary Agents

These are the entry points to the system. Start a conversation with any of them, and they'll orchestrate specialists as needed.

A

@agent-unleashed

Unrestricted general-purpose agent — plan, build, research, ship. Full tool access, minimal process. Use when you want maximum LLM judgment.

B

@builder

Builds features from PRDs or ad-hoc requests by orchestrating implementation agents

P

@planner

Refine draft PRDs and prepare them for implementation

T

@toolkit

Maintains the AI toolkit - agents, skills, templates, and scaffolds

U

@ui-test-full-app-auditor

Autonomous E2E test auditor that analyzes apps and generates comprehensive test coverage

View all agents →

The Human in the Loop (you)

You guide the direction. Agents handle the execution.

Shape the Vision

Work with @planner to refine requirements. You decide what gets built and how.

Review Progress

@builder implements your PRD. You verify, steer corrections, and approve commits.

Evolve the System

Improve agents over time via @toolkit. You approve changes that ripple across projects.

Explore human-agent collaboration →

The Process

The Agent Loop

From idea to shipped feature — and back again. Every feature follows this cycle.

P

Plan

@planner

🚀

Bootstrap

New project or existing codebase

📝

Draft PRD

Describe the feature

💬

Q&A

Clarify with lettered options

✅

Ready

PRD approved for build

B

Build

@builder → @developer

🎯

Claim

Lock PRD, create branch

⚡

Implement

loops

@developer writes code

🔍

Review

loops

@critic checks quality

🔄

Iterate

loops

Fix issues, repeat

🔄Repeat until all stories pass
T

Test

@tester → @critic

🧪

Unit Tests

loops

@tester adds coverage

🎭

E2E Tests

loops

End-to-end tests via @playwright-dev

🚦

Quality Gates

Typecheck, lint, build

📋

Human Test Script

Auto-generates QA checklist for manual testing

🔄Retry up to 3 times, then stop for human review
S

Ship

@builder → @felix

🔀

Create PR

Push and open PR

👁️

Watch CI

@felix monitors build

🎉

Merge

Ship to production

📦

Archive

PRD completed, cleanup

🐛

Bugs feed back into the loop

Discovered issues can be tracked as new PRDs

Learn about the Agent Loop →

Swipe to explore

66 Agents

Specialists on Call

When primary agents need focused expertise, specialists step in automatically.

29

Critics

Review code for correctness, security, style, and best practices.

17

Developers

Implement features across different languages and frameworks.

9

Testers

Write and run unit tests, integration tests, and E2E tests.

3

Orchestrators

Coordinate other agents and manage workflows.

8

Utilities

Debugging, docs, cleanup, and specialized tools.

Browse all agents →

73 Skills

Workflows That Load When Needed

Skills are specialized instruction sets for complex, multi-step work. Agents load them on-demand—situational expertise, not always-on overhead.

adhoc-workflow

Ad-hoc mode workflow for Builder

agent-audit

Audit AI toolkit agents for compliance and coverage gaps

agent-onboard

Onboard new or modified agents to ensure they are project-context aware and follow toolkit conventions

auth-config-check

Check authentication configuration before E2E tests

auth-generic

Generic/custom authentication implementation for any provider

auth-headless

Headless authentication using API calls and session injection

auth-nextauth-credentials

NextAuth credentials (email/password) authentication for AI agents

auth-supabase-otp

Supabase passwordless OTP authentication for AI agents

View all 73 skills →

Set Up in Minutes

Three steps. Then you're building.

1

Install opencode

A terminal-based AI coding tool. One command to install.

curl -fsSL https://opencode.ai/install.sh | bash
2

Connect a provider

Launch opencode and run /connect to link your AI provider (GitHub Copilot, Anthropic, OpenAI, etc.).

3

Install the toolkit

Add the agents, skills, and project templates.

curl -fsSL https://raw.githubusercontent.com/mdmagnuson-creator/yo-go/main/install.sh | bash
Full setup guide →