# Introduction

# Introduction

`@maestria/opencode` is an npm plugin for [OpenCode](https://opencode.ai) that transforms your AI coding environment into a disciplined engineering workstation. It bundles 8 specialized agents and global workflow rules — all installed with a single line of configuration.

## The Problem

Without a plugin like this, every OpenCode session starts from scratch. You need to:

- Manually configure agent prompts and permissions
- Remember to set context management rules
- Re-establish delegation patterns
- Hope the AI remembers its methodology

This is inconsistent, error-prone, and wastes context window on setup that should be automatic.

## The Approach

`@maestria/opencode` solves this with three layers:

| Layer      | What it does                                         | How                                                       |
| ---------- | ---------------------------------------------------- | --------------------------------------------------------- |
| **Agents** | 8 specialized subagents for different work phases    | Bundled `.md` files registered via OpenCode's config hook |
| **Rules**  | Cross-cutting directives injected into every session | `input.instructions` path to `rules/AGENTS.md`            |
| **Hooks**  | Self-wiring plugin architecture                      | OpenCode lifecycle hooks (`config`, `session.compacting`) |

## The 8 Agents

The plugin provides one manager agent and 7 specialist subagents:

| Agent           | Mode     | Role                                                                    |
| --------------- | -------- | ----------------------------------------------------------------------- |
| `@orchestrator` | all      | Manager — decomposes work, delegates to specialists, integrates results |
| `@adventurer`   | subagent | Codebase reconnaissance — maps unfamiliar code, traces call chains      |
| `@architect`    | subagent | Architecture decisions — evaluates options with weighted criteria       |
| `@builder`      | subagent | Focused implementation — one atomic task per invocation                 |
| `@diagnose`     | subagent | Systematic debugging — 6-step regression tracing                        |
| `@planner`      | subagent | Implementation plans — phased milestones with verification              |
| `@reviewer`     | subagent | Code review — 7-dimension quality checklist                             |
| `@writer`       | subagent | Documentation — structured patterns for any document type               |

## Next Steps

- [Install the plugin](/opencode/getting-started/installation/) — Add it to your OpenCode config
- [Quick Start](/opencode/getting-started/quick-start/) — Your first session with the plugin
- [Browse the Agents](/opencode/agents/) — Detailed documentation for each agent