Skip to content
Systematic Structured Engineering Workflows for OpenCode

Structured Engineering Workflows for OpenCode

@fro.bot/systematic

Your AI writes code.
Systematic adds engineering discipline.

Systematic brings structured workflows to OpenCode. Brainstorm, plan, implement, and review with bundled skills and agents.

Install the PluginView on GitHub

See the Difference

Most AI coding sessions produce code. Systematic produces a process. Brainstorm before planning, plan before building, review before shipping.

Walk through the workflow

What You Get

Bundled Skills

Structured workflows for every phase: brainstorm, plan, work, review, compound, TDD, browser testing, and more. Each skill encodes a repeatable engineering process, not just a prompt.

Specialized Agents

Purpose-built subagents for architecture, security, performance, design, and code review. They run inside OpenCode and inherit your model, requiring no extra configuration.

Bootstrap Injection

Drop one line into your OpenCode config. Skills and agents register automatically via plugin hooks. Nothing to wire up, no scaffolding commands to run.

Typed Config

Plugin configuration is fully typed. Your editor catches mistakes before you run anything, ensuring your setup is correct from the start.

Content-Integrity CI

A built-in gate enforces structural rules on bundled agents. No model field in subagent frontmatter, no banned patterns. Catches authoring mistakes before they ship.

Engineering Effort That Compounds

Right now, every new task takes roughly the same effort as the last one. Systematic changes that.

When you run /ce:compound after a solved problem, you’re writing the solution into the system permanently. The next time that problem comes up, or a close relative of it, the agent already has the pattern. The tenth task is faster than the first. The hundredth can run in parallel.

This isn’t a feature. It’s the entire design goal.

Read the philosophy

A Loop You Can Trust

Ad-hoc AI sessions produce ad-hoc results. Systematic structures the session into stages you can audit.

The loop is: brainstorm → plan → work → review → compound. Each stage has a skill. Each skill has a defined scope and exit condition. You know what the agent is supposed to do at each step, and you can verify it did.

No hidden prompts. No global context injection. You invoke skills explicitly when you want them.

Quick Start

OpenCode Plugin (Recommended)

OpenCode, full runtime

Add Systematic to your configuration:

{
  "plugins": ["@fro.bot/systematic@latest"]
}

Then start your first task:

Terminal window
/ce:brainstorm "add dark mode toggle"

CLI Installer

Any harness, skills only

Install Markdown skills directly into your repo:

Terminal window
npx skills add marcusrbrown/systematic

Skills and agents are available immediately. No API keys, no accounts, no telemetry.

What Systematic Is Not

  • Not a model wrapper or prompt pack. Systematic doesn’t touch your model selection or inject hidden system prompts into every message. It registers structured skills you invoke explicitly.

  • Not another chat UI. It runs inside OpenCode. You keep your existing setup.

  • Not SaaS or telemetry. Everything is local-first markdown assets bundled with the npm package. Nothing phones home.

  • Not magic autopilot. The AI still does the work. Systematic gives it a process to follow, the same process a disciplined engineer would use. You stay in the loop.

Privacy