Skip to content

README.en.md

Latest commit

 

History

History
213 lines (162 loc) · 8.77 KB

README.en.md

File metadata and controls

213 lines (162 loc) · 8.77 KB

Opencode-agents

A ready-to-use collection of OpenCode agent configurations: 2 primary agents and 25 specialized subagents covering architecture, code review, debugging, testing, frontend development, database engineering, DevOps, and more.

Project Structure

opencode-agents/
├── opencode.jsonc          # OpenCode global config (permissions, default agent)
├── agents/
│   ├── Zero.md             # Primary — rapid prototyping (mimo-v2.5, multimodal)
│   ├── Erribaba.md         # Primary — production coding (mimo-v2.5-pro)
│   ├── api-designer.md     # API endpoint design, OpenAPI specs
│   ├── architect.md        # System architecture, module planning
│   ├── code-generator.md   # High-quality code generation, algorithms
│   ├── db-engineer.md      # Database schema, migrations, SQL optimization
│   ├── debugger.md         # Systematic debugging, root cause analysis
│   ├── devops.md           # Docker, CI/CD, Kubernetes, IaC
│   ├── doc-writer.md       # Technical docs, README, API references
│   ├── e2e-tester.md       # Playwright/Cypress end-to-end tests
│   ├── executor.md         # Command execution, test running, builds
│   ├── frontend-dev.md     # React/Vue/Svelte component development
│   ├── frontend-reviewer.md# Frontend review, WCAG 2.1 AA accessibility
│   ├── git-assistant.md    # Commit messages, branch naming, PR descriptions
│   ├── migration.md        # DB migration, framework upgrades, stack changes
│   ├── perf-optimizer.md   # Performance profiling and optimization
│   ├── project-manager.md  # Requirements breakdown, sprint planning
│   ├── refactorer.md       # Code refactoring, duplication removal
│   ├── research.md         # Documentation lookup, tech research
│   ├── reviewer.md         # Backend code review (logic/security/performance)
│   ├── security-auditor.md # OWASP Top 10 security audit
│   ├── software-engineer.md# Full-stack end-to-end implementation
│   ├── test-writer.md      # Unit, integration, and edge-case tests
│   ├── ui-designer.md      # CSS/Tailwind, responsive layouts, animations
│   ├── validator.md        # Final validation (build/tests/type check)
│   ├── explore.md          # Codebase exploration, structure analysis
│   └── vision-dev.md       # Design screenshot analysis, image-to-code
├── AGENTS.md               # Contributing guide & agent authoring spec
├── CHANGELOG.md            # Version history
├── README.md               # Chinese
├── README.en.md            # English (this file)
└── LICENSE                 # MIT

Agent Overview

Primary Agents

File Model Purpose
Zero.md opencode-go/mimo-v2.5 Rapid prototyping — multimodal input, visual analysis, lightweight tasks
Erribaba.md opencode-go/mimo-v2.5-pro Production coding — complex algorithms, deep analysis, perf-critical code

Both primary agents share the same subagent orchestration system. The difference is model capability and use case:

  • Zero: Fast, supports image input, ideal for prototyping and simple tasks
  • Erribaba: Stronger coding ability, ideal for production code and complex engineering

Subagents

Code Quality & Review

File Model Responsibility Write
reviewer.md deepseek-v4-pro Backend review (logic/security/performance/types) ✗
frontend-reviewer.md mimo-v2.5 Frontend review (components/perf/WCAG 2.1 AA/CSS) ✗
security-auditor.md deepseek-v4-pro OWASP Top 10 security audit ✗
validator.md minimax-m2.7 Final validation (build/tests/types/lint) ✗

Testing

File Model Responsibility Write
test-writer.md mimo-v2.5-pro Unit/integration/edge-case tests (AAA pattern) ✓
e2e-tester.md mimo-v2.5-pro Playwright/Cypress end-to-end tests ✓

Debugging & Optimization

File Model Responsibility Write
debugger.md deepseek-v4-pro Systematic debugging, root cause, minimal fix ✓
perf-optimizer.md deepseek-v4-pro Performance profiling, bottleneck identification ✗
refactorer.md mimo-v2.5-pro Code refactoring, extract functions, DRY ✓

Architecture & Design

File Model Responsibility Write
architect.md glm-5.1 System architecture, module planning, tech selection ✗
api-designer.md mimo-v2.5-pro API endpoint design, OpenAPI specs, error codes ✓

Implementation

File Model Responsibility Write
code-generator.md mimo-v2.5-pro High-quality code gen, algorithms, bug fixes ✓
software-engineer.md mimo-v2.5 Full-stack end-to-end implementation ✓
frontend-dev.md kimi-k2.6 React/Vue/Svelte component development ✓
ui-designer.md kimi-k2.6 CSS/Tailwind, responsive layouts, animations ✓
vision-dev.md mimo-v2.5 Design analysis, screenshot-to-code (multimodal) ✓

Data & Infrastructure

File Model Responsibility Write
db-engineer.md deepseek-v4-pro Schema design, migrations, SQL optimization ✓
devops.md mimo-v2.5-pro Docker, CI/CD, Kubernetes, IaC ✓
migration.md deepseek-v4-pro DB migration, framework upgrades, stack changes ✓

Documentation & Planning

File Model Responsibility Write
doc-writer.md qwen3.6-plus Technical docs, README, API references ✓
project-manager.md qwen3.6-plus Requirements breakdown, sprint planning ✓
git-assistant.md mimo-v2.5 Commit messages, branch naming, PR descriptions ✗
research.md qwen3.6-plus Documentation lookup, tech research ✗

Execution

File Model Responsibility Write
executor.md minimax-m2.7 Command execution, test running, builds ✓

Exploration

File Model Responsibility Write
explore.md qwen3.6-plus Codebase search, project structure analysis ✗

Usage

1. Copy Agent Files

Copy the agents/ directory to your OpenCode config location:

# Global config
cp -r agents/ ~/.config/opencode/agents/

# Or project-level config
cp -r agents/ <your-project>/.opencode/agents/

2. Configure opencode.jsonc

Copy opencode.jsonc to your OpenCode config directory:

cp opencode.jsonc ~/.config/opencode/opencode.jsonc

This file defines:

  • default_agent — which primary agent to use (Zero or Erribaba)
  • permission — tool permissions (file ops, command execution, network access, etc.)

3. Start Using

After launching OpenCode, the primary agent automatically delegates tasks to the appropriate subagent. You can also explicitly request a specific subagent in your conversation.

File Format

Each agent file consists of YAML frontmatter + Markdown body:

---
description: Chinese description (used by OpenCode for trigger matching)
mode: primary | subagent
model: provider/model-name
temperature: 0.0-1.0
tools:              # optional — restrict tool access
  write: false
  edit: false
  bash: false
---

<System prompt>

Field Reference

Field Required Description
description Yes Chinese text — OpenCode uses this to decide when to invoke the agent
mode Yes primary (top-level session) or subagent (invoked via delegation only)
model Yes Model identifier in provider/model-name format
temperature Yes Generation temperature, 0.0-1.0
tools No Tool permissions; omitted keys default to allowed; set false to deny

Models Used

Model ID Purpose
opencode-go/mimo-v2.5 Multimodal, rapid prototyping, general tasks
opencode-go/mimo-v2.5-pro Production coding, complex algorithms
opencode-go/deepseek-v4-pro Code review, security audit, database, debugging
opencode-go/kimi-k2.6 Frontend development, UI design
opencode-go/qwen3.6-plus Documentation, research, project management
opencode-go/glm-5.1 Architecture design
opencode-go/minimax-m2.7 Command execution, result validation

Contributing

  1. Fork the repository
  2. Create a Feat_xxx branch
  3. Commit your code
  4. Create a Pull Request

For detailed agent authoring guidelines, see AGENTS.md.

License

MIT