mirror of
https://github.com/IgorWarzocha/Opencode-Workflows.git
synced 2026-09-14 16:22:53 +08:00
OpenCode Agent
01ee76f017
- Move top-level agent directories under agents/ - Remove deprecated skill permission warnings - Update AGENTS.md to reflect new repository structure
69 lines
2.8 KiB
Markdown
69 lines
2.8 KiB
Markdown
# OpenCode ConFIGurator Pack
|
|
|
|
A meta-configuration system for [OpenCode](https://opencode.ai). Ask questions about plugins, create custom commands, build agents, design skills — all through natural conversation.
|
|
|
|
## The ConFIGurator Agent
|
|
|
|
Your OpenCode power user on demand. It knows the official documentation and will fetch the latest when needed. It can validate your `opencode.json`, explain options, and fix issues. When hunting for plugins, it searches the web and npm, then documents discoveries for future sessions.
|
|
|
|
## The Skills
|
|
|
|
The agent draws on seven specialized skills:
|
|
|
|
| Skill | What it does |
|
|
|-------|--------------|
|
|
| **plugin-installer** | Find and install community plugins, maintain a local catalog of discoveries |
|
|
| **opencode-config** | Edit your `opencode.json` with guided setup for models, permissions, and providers |
|
|
| **command-creator** | Build custom `/slash` commands through interactive Q&A |
|
|
| **skill-creator** | Scaffold new skills with proper structure, scripts, and references |
|
|
| **agent-architect** | Design agents with research-backed prompt engineering patterns |
|
|
| **mcp-installer** | Find, install, and configure Model Context Protocol (MCP) servers |
|
|
| **model-researcher** | Research and configure new/custom AI models not yet in models.dev |
|
|
|
|
## The Commands
|
|
|
|
| Command | What it does |
|
|
|---------|-------------|
|
|
| **refactor-rfc-xml** | Convert markdown files to RFC 2119 + XML tag structure |
|
|
|
|
## Installation
|
|
|
|
Drop the `skill/`, `agent/`, and `command/` folders into `~/.config/opencode/`.
|
|
|
|
> [!NOTE]
|
|
> The opencode-configurator agent has **7 skills enabled by default**: `plugin-installer`, `opencode-config`, `command-creator`, `skill-creator`, `agent-architect`, `mcp-installer`, and `model-researcher`. All other skills are blocked with `"*": "deny"`. This allows the configurator to dynamically load the appropriate skill for each task without context pollution.
|
|
|
|
### Skill Configuration
|
|
|
|
The agent's YAML frontmatter controls skill access:
|
|
|
|
```yaml
|
|
permission:
|
|
skill:
|
|
plugin-installer: allow
|
|
opencode-config: allow
|
|
command-creator: allow
|
|
skill-creator: allow
|
|
agent-architect: allow
|
|
mcp-installer: allow
|
|
model-researcher: allow
|
|
'*': deny
|
|
```
|
|
|
|
## Usage
|
|
|
|
Just ask naturally:
|
|
|
|
- *"Is there a plugin for reducing token usage?"*
|
|
- *"Set up permissions so destructive commands require approval"*
|
|
- *"Create a /review command that analyzes code without making changes"*
|
|
- *"I want to build a skill for working with our internal API"*
|
|
- *"Make me an agent for deployment that can only run git and npm commands"*
|
|
- *"Find an MCP server for filesystem access"*
|
|
|
|
The configurator figures out which skill to load and walks you through it.
|
|
|
|
## Contributing
|
|
|
|
Found a useful plugin that isn't in the catalog? The plugin-installer skill can document it for future sessions. Discovered patterns worth sharing? Submit a PR.
|