opencode-enhancer-plugin

v1.5.0

Published

a month ago

Universal Technical Architect & Prompt Enhancer with Todo Enforcer and Strategos Planning Mode

0High
0Medium
0Low

maansen

opencode plugin enhancer ultraplan prompt context ai agent openai anthropic gemini

🚀 OpenCode Enhancer Plugin

Universal Technical Architect & Prompt Enhancer - Transform vague ideas into precise, context-aware engineering specifications.

✨ What is Enhancer?

The Enhancer is a third primary agent mode for OpenCode that sits alongside Plan and Build. It acts as your technical architect, analyzing your codebase and generating detailed, actionable prompts that you can execute in Build mode.

🎯 Key Features

🔍 Intelligent Context Analysis - Automatically explores your project structure, tech stack, and relevant files
🧠 Intent Classification - Categorizes requests as FIX, FEAT, REFACTOR, TEST, or EXPLAIN
🔒 Todo Enforcer System - Tracks TODOs and prevents stopping until all tasks are completed (for ultraplan and strategos agents)
🎯 Strategos Mode - Strategic planning with interview mode for complex tasks
📚 Context7 Integration - Conditionally includes use context7 when documentation-heavy frameworks are detected
🎨 Style-Aware - Mimics your project's coding style and patterns
⚡ One-Click Workflow - Generates copy-pasteable prompts ready for Build mode

🚀 Quick Start

Installation

From npm (Recommended)

Add to your OpenCode config (~/.config/opencode/opencode.json):

{
  "$schema": "https://opencode.ai/config.json",
  "plugin": [
    "opencode-enhancer-plugin@latest"
  ]
}

Restart OpenCode and press Tab to cycle through modes:

Plan → Build → Enhancer → Plan

From Source (Development)

Clone this repository:

git clone https://github.com/MaansenV/opencode-enhancer-plugin.git
cd opencode-enhancer-plugin

Install dependencies and build:
```
npm install
npm run build
```

Add to your OpenCode config using the local path:

{
  "plugin": [
    "file:///path/to/opencode-enhancer-plugin/dist/index.js"
  ]
}

Version Management

Latest version: "opencode-enhancer-plugin@latest"
Current version: "[email protected]"
Specific version: "[email protected]"
Update: OpenCode auto-updates plugins on startup, or run opencode --update-plugins

Uninstall

Remove from your opencode.json:

{
  "plugin": []
}

📖 How It Works

┌─────────────┐     ┌──────────────┐     ┌─────────────┐
│   User      │────▶│  Enhancer    │────▶│   Build     │
│  Request    │     │   Agent      │     │   Mode      │
└─────────────┘     └──────────────┘     └─────────────┘
                           │
                           ▼
                    ┌──────────────┐
                    │  Explore-    │
                    │   Context    │
                    │  (Subagent)  │
                    └──────────────┘

Workflow

Select Enhancer Mode (Tab to cycle)
Describe your goal - "Fix the login bug" or "Add user authentication"
Enhancer analyzes - Calls explore-context to map your codebase
Receive enhanced prompt - Structured markdown with context, instructions, and technical requirements
Copy & Execute - Switch to Build mode and run the generated prompt

📝 Changelog

v1.5.0 - Specialized Subagents

📚 Librarian - Documentation and research specialist with Context7 and Code Search
🔮 Oracle - Debugging and code architecture specialist
🏗️ Architect - Dependencies and system design specialist
✅ Reviewer - Plan validation specialist
🎤 Interviewer - Requirements clarification specialist
Optimized 4+1 subagent structure (4 specialists + 1 interviewer)
Simplified, focused prompts for better performance

v1.4.0 - Todo Enforcer & Strategos Mode

🔒 Todo Enforcer System - Tracks TODOs and prevents stopping until all tasks are completed (for ultraplan and strategos agents)
🎯 Strategos Mode - New primary agent with strategic planning and interview mode for complex multi-phase tasks
strategos-interviewer subagent for stakeholder requirement gathering
Enhanced hooks for session management and TODO tracking
Auto-activation of strategos mode on keyword detection

v1.2.0 - New Agents & Model Configuration

ultraplan & ask primary agents
review-plan subagent for critical analysis
Enhanced model configuration with environment variables

v1.0.0 - Initial Release

enhancer primary agent
explore-context subagent for project analysis
Intent classification and Context7 integration
Style-aware prompt generation

🎨 Example Output

# Task: Fix Authentication Token Validation

## Context
Detected **React + TypeScript + Express** project. Authentication logic located in `src/auth/middleware.ts` and `src/utils/token.ts`.

## Instructions
use context7
1. Modify `src/auth/middleware.ts`:
   - Locate function `validateToken` (line 45)
   - Add null check for `req.headers.authorization`
   - Implement JWT verification using existing `verify()` pattern
2. Update `src/utils/token.ts`:
   - Add error handling for expired tokens
   - Return structured error object instead of throwing
3. Testing:
   - Run `npm test` to verify auth flow
   - Check edge cases: missing token, expired token, malformed token

## Technical Requirements
- **Style**: Follow existing async/await patterns
- **Error Handling**: Use `try/catch` with custom AuthError class
- **TypeScript**: Maintain strict typing, add interfaces for token payload

🛠️ Architecture

Components

Primary Agents (User-Facing)

Activated via Tab key in OpenCode:

| Agent | Color | Steps | Description | |-------|-------|-------|-------------| | enhancer | Purple (#9C27B0) | 15 | Universal Technical Architect - analyzes intent and generates executable prompts | | ultraplan | Orange (#FF5722) | 20 | Iterative Planner with Todo Enforcement - creates plans through parallel subagent analysis and review loops | | ask | Blue (#2196F3) | 15 | Research Assistant - answers codebase questions via multi-source context gathering | | strategos | Deep Orange (#BF360C) | 30 | Strategic Planner with Interview Mode - handles complex multi-phase tasks with strategic planning and stakeholder interviews |

Subagents (Background Workers)

Called automatically by primary agents via task tool:

| Subagent | Called By | Description | |----------|-----------|-------------| | librarian | All primary agents | Documentation and codebase research specialist (Context7 + codesearch) | | oracle | All primary agents | Debugging and code architecture specialist | | architect | ultraplan, strategos, ask | Dependencies and system design specialist | | reviewer | ultraplan, strategos | Plan validation and critical analysis specialist | | interviewer | strategos only | Requirements clarification and stakeholder interview specialist |

Supported Tech Stacks

Web: React, Vue, Angular, Svelte, Next.js, Nuxt, Vite
Backend: Express, Django, Flask, FastAPI, Rails, Laravel, Spring
Game Engines: Unity, Godot, Unreal Engine
Data/ML: TensorFlow, PyTorch, NumPy, Pandas
Infrastructure: Docker, Kubernetes, Terraform
And more...

⚙️ Configuration

Agent Configuration

The plugin automatically registers agents via the config hook with configurable models:

// Enhancer (Primary Agent) - Model configurable via opencode.json or env vars
{
  mode: "primary",
  model: "opencode/kimi-k2.5-free",  // Default, can be overridden
  description: "Universal Technical Architect & Prompt Enhancer",
  color: "#9C27B0",
  steps: 15,
  tools: { task: true, read: true }
}

// Explore-Context (Subagent) - Model configurable via opencode.json or env vars
{
  mode: "subagent",
  hidden: true,
  model: "opencode/kimi-k2.5-free",  // Default, can be overridden
  tools: { list: true, read: true, grep: true, bash: true }
}

Model Configuration

You can configure custom models for Enhancer agents. This is useful for:

Using more powerful models for complex planning tasks
Saving costs by using simpler models for subagents
Testing specific models for certain task types

Configuration Methods (Priority: Highest → Lowest)

1. OpenCode Agent Config (Recommended)

Configure models directly in your opencode.json:

{
  "agent": {
    "strategos": {
      "model": "anthropic/claude-sonnet-4-5"
    },
    "oracle": {
      "model": "anthropic/claude-sonnet-4-5"
    },
    "librarian": {
      "model": "opencode/kimi-k2.5-free"
    }
  }
}

2. Environment Variables

Use environment variables for CI/CD or temporary changes:

# All Subagents
export ENHANCER_MODEL_SUBAGENT="opencode/kimi-k2.5-free"

# All Primary Agents  
export ENHANCER_MODEL_PRIMARY="anthropic/claude-sonnet-4-5"

# Specific Agent (highest priority among env vars)
export ENHANCER_MODEL_STRATEGOS="anthropic/claude-opus-4"
export ENHANCER_MODEL_ORACLE="anthropic/claude-sonnet-4-5"

Model Recommendations by Agent Type

Primary Agents (Use stronger models for complex reasoning):

| Agent | Recommended | Minimum | Purpose | |-------|-----------|---------|---------| | strategos | anthropic/claude-sonnet-4-5 | opencode/kimi-k2.5-free | Strategic planning with interview mode | | ultraplan | anthropic/claude-sonnet-4-5 | opencode/kimi-k2.5-free | Planning with todo enforcement | | enhancer | opencode/kimi-k2.5-free | opencode/kimi-k2.5-free | Prompt enhancement & analysis | | ask | opencode/kimi-k2.5-free | opencode/kimi-k2.5-free | Q&A and research |

Subagents (Lightweight models sufficient for information gathering):

| Subagent | Recommended | Minimum | Purpose | |----------|-------------|---------|---------| | librarian | opencode/kimi-k2.5-free | opencode/kimi-k2.5-free | Documentation & research (Context7 + codesearch) | | oracle | anthropic/claude-sonnet-4-5 | opencode/kimi-k2.5-free | Debugging & code architecture | | architect | opencode/kimi-k2.5-free | opencode/kimi-k2.5-free | Dependencies & system design | | reviewer | opencode/kimi-k2.5-free | opencode/kimi-k2.5-free | Plan validation | | interviewer | anthropic/claude-sonnet-4-5 | opencode/kimi-k2.5-free | Requirements clarification |

Validation

The plugin validates configured models against known providers:

opencode/*
anthropic/*
openai/*
google/*
mistral/*
cohere/*
ollama/*

Invalid models automatically fall back to the default (opencode/kimi-k2.5-free).

🎨 Ready-to-Use Templates

Siehe MODEL-TEMPLATES.md für vorgefertigte Konfigurationen:

🚀 Performance First - Beste Ergebnisse mit Claude
💰 Budget First - Kosten sparen mit Kimi
⚡ Hybrid - Intelligente Aufteilung nach Aufgabe
🔧 Custom Models - OpenAI, Google, etc.

Quick-Start:

# Hybrid Template (empfohlen für den Anfang)
curl -s https://raw.githubusercontent.com/MaansenV/opencode-enhancer-plugin/main/MODEL-TEMPLATES.md | grep -A 50 "Template 3: Hybrid"

Context7 Integration

The plugin automatically detects libraries that benefit from Context7:

// Libraries triggering "use context7"
["react", "django", "unity", "tensorflow", "docker", ...]

🔧 Development

Project Structure

opencode-enhancer-plugin/
├── src/
│   └── index.ts          # Source TypeScript
├── dist/                 # Compiled output (published to npm)
│   ├── index.js
│   └── index.d.ts
├── package.json          # Package metadata
├── tsconfig.json         # TypeScript config
└── README.md             # This file

Hooks

config - Registers agents on startup
tool.execute.before - Logs subagent calls
message.updated - Appends Build mode hint
message.completed - Extracts and stores TODOs
stop.requested - Prevents stop when TODOs are open
session.start - Manages todo store lifecycle
session.end - Manages todo store lifecycle
user.prompt.submitted - Auto-activates strategos mode on keywords

🤝 Contributing

Contributions are welcome! Please feel free to submit a Pull Request.

Ideas for Contributions

Add more tech stack detectors
Improve Context7 heuristics
Add support for additional intent types
Enhance prompt templates

📄 License

MIT License - see LICENSE file for details.

🙏 Acknowledgments

Built for OpenCode
Inspired by the need for better context-aware AI assistance
Thanks to the OpenCode community for feedback and ideas

⬆ Back to Top

Made with ❤️ for the OpenCode community