/ 目錄 / 演練場 / Flow Next
● 社群 gmickel ⚡ 即開即用

Flow Next

作者 gmickel · gmickel/flow-next

Plan-first orchestration: every non-trivial ask becomes a checkable plan you can edit before any code is written.

Flow Next slots between you and the agent: it prompts for a structured plan first, you approve, then execution. Works across Claude Code, OpenAI Codex, and Factory Droid.

為什麼要用

核心特性

即時演示

實際使用效果

就緒

安裝

選擇你的客戶端

~/Library/Application Support/Claude/claude_desktop_config.json  · Windows: %APPDATA%\Claude\claude_desktop_config.json
{
  "mcpServers": {
    "flow-next-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/gmickel/flow-next"
      ],
      "_inferred": true
    }
  }
}

開啟 Claude Desktop → Settings → Developer → Edit Config。儲存後重啟應用。

~/.cursor/mcp.json · .cursor/mcp.json
{
  "mcpServers": {
    "flow-next-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/gmickel/flow-next"
      ],
      "_inferred": true
    }
  }
}

Cursor 使用與 Claude Desktop 相同的 mcpServers 格式。專案級設定優先於全域。

VS Code → Cline → MCP Servers → Edit
{
  "mcpServers": {
    "flow-next-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/gmickel/flow-next"
      ],
      "_inferred": true
    }
  }
}

點擊 Cline 側欄中的 MCP Servers 圖示,然後選 "Edit Configuration"。

~/.codeium/windsurf/mcp_config.json
{
  "mcpServers": {
    "flow-next-skill": {
      "command": "git",
      "args": [
        "clone",
        "https://github.com/gmickel/flow-next"
      ],
      "_inferred": true
    }
  }
}

格式與 Claude Desktop 相同。重啟 Windsurf 生效。

~/.continue/config.json
{
  "mcpServers": [
    {
      "name": "flow-next-skill",
      "command": "git",
      "args": [
        "clone",
        "https://github.com/gmickel/flow-next"
      ]
    }
  ]
}

Continue 使用伺服器物件陣列,而非映射。

~/.config/zed/settings.json
{
  "context_servers": {
    "flow-next-skill": {
      "command": {
        "path": "git",
        "args": [
          "clone",
          "https://github.com/gmickel/flow-next"
        ]
      }
    }
  }
}

加入 context_servers。Zed 儲存後熱重載。

claude mcp add flow-next-skill -- git clone https://github.com/gmickel/flow-next

一行命令搞定。用 claude mcp list 驗證,claude mcp remove 移除。

使用場景

實戰用法: Flow Next

Plan a feature before coding

👤 Claude Code users ⏱ ~15 min intermediate

何時使用: You want to align on approach before agents start writing.

步驟
  1. Trigger
    Invoke flow-next with your feature description.✓ 已複製
    → Plan emitted
  2. Edit
    Tweak / reorder / drop steps.✓ 已複製
    → Approved plan
  3. Execute
    Run the plan; agent ticks steps as they complete.✓ 已複製
    → Feature built per plan

結果: Reviewed plan executed without surprise scope creep.

組合

與其他 MCP 搭配,撬動十倍槓桿

flow-next-skill + github

Attach the plan to the PR for reviewer context

Combine flow-next-skill with github: Attach the plan to the PR for reviewer context✓ 已複製

工具

此 MCP 暴露的能力

工具輸入參數何時呼叫成本
plan / approve / execute (see docs) Three-phase command surface 1 call

成本與限制

運行它的成本

API 配額
N/A
每次呼叫 Token 數
Plans add ~1 plan-sized round-trip
費用
Free
提示
Save plans alongside the PR — great for retros

安全

權限、密鑰、影響範圍

最小權限: fs-read fs-write
憑證儲存: None
資料出站: None

故障排查

常見錯誤與修復

Plan too vague

Ask for "concrete file paths" in the brief

替代方案

Flow Next 對比其他方案

替代方案何時用它替代權衡
Direct executionTrivial changeRiskier on bigger work

更多

資源

📖 閱讀 GitHub 上的官方 README

🐙 查看未解決的 issue

🔍 瀏覽全部 400+ MCP 伺服器和 Skills