erne 初始化
- 作者仓库星标 36
- 许可证 MIT
- 作者仓库 everything-react-native-expo
Initialize ERNE
ERNE is an AI agent harness for React Native and Expo. Setting it up involves choosing a few preferences and then running a single CLI command. The reason you need to ask the user before running is that the CLI's interactive prompts don't work reliably in Claude Code — so you handle the interaction here in conversation instead.
Step 1: Ask preferences
Before running anything, ask the user two things in a single message:
Profile: ERNE uses hook profiles to control how much code quality checking happens automatically. Ask which one they'd like:
- minimal — Almost no automated checks. Good for quick prototyping or when you just want the agents without guardrails.
- standard — Catches common issues (formatting, console.logs, platform-specific bugs) without slowing things down. This is what most people pick.
- strict — Adds security scanning, accessibility checks, and test gates. Good for production apps or teams.
MCP servers: ERNE can configure agent-device (controls iOS Simulator and Android Emulator for screenshots, taps, navigation) and GitHub integration. Ask if they want these set up now or later.
Wait for the user to respond before continuing.
Step 2: Run the init command
Once you have their preferences, run this in Bash:
npx erne-universal@latest init --yes --profile <profile> [--no-mcp]Replace <profile> with their choice (minimal, standard, or strict). Add --no-mcp only if they said no to MCP servers.
The --yes flag is important — it tells the CLI to skip its own interactive prompts since you already gathered the preferences.
Step 3: Launch the dashboard
After init completes, start the dashboard:
npx erne-universal dashboard &The dashboard runs in the background. Read the output to find which port it started on (usually 3333, but it auto-selects a free port if 3333 is taken).
If the dashboard needs to install dependencies on first run (takes about 2 minutes), tell the user it's installing and will open in the browser when ready.
Step 4: Tell the user what happened
Summarize in a clear message:
- What was set up (agents, hooks, rules, skills, MCP servers)
- The dashboard URL (e.g., http://localhost:3333)
- That they need to restart this Claude Code session for MCP servers and hooks to activate
- A few commands they can try after restart:
/erne-plan— plan a new feature/erne-perf— profile performance issues/erne-doctor— check project health/erne-code-review— review code quality- Type
/erne-and press Tab to see all commands
- 流狐分类
- AI 智能
- 作者声明 Agent
- 未找到明确声明；不据此推断已兼容或已测试
- 静态检查
- 94 / 100 · 启发式扫描，不代表运行安全
- 作者 / 版本 / 许可
- @JubaKitiashvili · MIT
- 流狐 Token 估算
- 低消耗
- 流狐接入估算
- 即装即用
- 是否需要外部 API Key
- 未发现要求
- 检测到的系统要求
- macOS · Linux · Windows
- 底层运行要求
- 未声明
- 检测到的文件与系统行为
-
- 只读
- 检测到的网络行为
- 仅限本地
- 安装命令数
- 无（仅作为资料）
档案由构建时根据 SKILL.md 与安装命令自动衍生，可能与作者实际意图存在差异。
需要注意： 未限定 allowed-tools，默认拥有全部工具权限。
作者没有在当前 SKILL.md 中定义固定输出样例。Before running anything, ask the user two things in a single message: Profile: ERNE uses hook profiles to control how much code quality checking happens automatically. Ask which one they'd like: minimal — Almost no automated checks. Good for quick prototyping…
Once you have their preferences, run this in Bash: Replace <profile> with their choice (minimal, standard, or strict). Add --no-mcp only if they said no to MCP servers. The --yes flag is important — it tells the CLI to skip its own interactive prompts since…
After init completes, start the dashboard: The dashboard runs in the background. Read the output to find which port it started on (usually 3333, but it auto-selects a free port if 3333 is taken). If the dashboard needs to install dependencies on first run…
Summarize in a clear message: What was set up (agents, hooks, rules, skills, MCP servers) The dashboard URL (e.g., http://localhost:3333)
# Initialize ERNE
ERNE is an AI agent harness for React Native and Expo. Setting it up involves choosing a few preferences and then running a single CLI command. The reason you need to ask the user before running is that the CLI's interactive prompts don't work reliably in Claude Code — so you handle the interaction here in conversation instead.
## Step 1: Ask preferences
Before running anything, ask the user two things in a single message:
**Profile:** ERNE uses hook profiles to control how much code quality checking happens automatically. Ask which one they'd like:
- **minimal** — Almost no automated checks. Good for quick prototyping or when you just want the agents without guardrails.
- **standard** — Catches common issues (formatting, console.logs, platform-specific bugs) without slowing things down. This is what most people pick.
- **strict** — Adds security scanning, accessibility checks, and test gates. Good for production apps or teams.
**MCP servers:** ERNE can configure agent-device (controls iOS Simulator and Android Emulator for screenshots, taps, navigation) and GitHub integration. Ask if they want these set up now or later.
Wait for the user to respond before continuing.
## Step 2: Run the init command
Once you have their preferences, run this in Bash:
```bash
npx erne-universal@latest init --yes --profile <profile> [--no-mcp]
```
Replace `<profile>` with their choice (minimal, standard, or strict). Add `--no-mcp` only if they said no to MCP servers.
The `--yes` flag is important — it tells the CLI to skip its own interactive prompts since you already gathered the preferences.
## Step 3: Launch the dashboard
After init completes, start the dashboard:
```bash
npx erne-universal dashboard &
```
…作者原文负责流程事实；流狐只索引当前章节、要点、文件与命令。
章节 -> Step 1: Ask preferences → Step 2: Run the init command → Step 3: Launch the dashboard → Step 4: Tell the user what happened
要点 -> Profile · minimal · standard · strict · MCP servers · restart this Claude Code session
文件/命令 -> <profile> · --no-mcp · --yes · /erne-plan · /erne-perf · /erne-doctor · /erne-code-review · /erne-
内容 SHA-256 -> 226795d71355
方法与流程
适用与边界
原文中的明确线索
<profile>、--no-mcp、--yes、/erne-plan、/erne-perf、/erne-doctor、/erne-code-review、/erne-