Token导航 LogoToken导航

Agent Skill · 开发

dev-terminal(开发终端)

dev-terminal 用于处理 GitHub 仓库、Issue、Pull Request 和代码协作信息,适合在 Codex、Claude、Cursor、Gemini CLI 中需要围绕仓库状态、代码变更或协作事项进行整理时使用。可结合来源仓库、安装命令和原始 README 继续核验具体用法。安装前建议确认权限范围、维护状态,以及是否会触发联网、命令执行或文件读写。

dev-terminal 是开发类 Agent Skill,可用宿主为 Codex、Claude、Cursor、Gemini CLI,通过GitHub 仓库获取,权限为接触敏感数据;许可证为 MIT,第三方扫描有未通过项,按安装量在开发类的 21,746 个 Skill 里排第 13,087。

总安装 1,018周安装 42GitHub Stars 1下载 333核验 2026-05-01

安装说明

本站只整理中文说明,不托管安装包,也不代用户安装。
GitHub
许可证
MIT

安装方式

通过对话安装

复制提示词发给支持本地命令或 Skills 的 AI 助手,先确认命令和权限,再让它执行。

请帮我安装这个 Agent Skill:dev-terminal(开发终端)
来源仓库:https://github.com/parkerhancock/dev-terminal
仓库路径:skills/dev-terminal
安装命令:
npx skills add https://github.com/parkerhancock/dev-terminal --skill dev-terminal
安装前请先检查当前环境是否支持对应 CLI,并向我确认将要执行的命令、安装目录、联网范围和文件读写权限;确认后再执行。

命令行安装

复制命令到本机终端执行。不同来源提供的安装方式可能略有差异;本站展示可直接复制的安装命令,安装前请核对来源页面。

skills.shnpx skills
npx skills add https://github.com/parkerhancock/dev-terminal --skill dev-terminal

这个 Skill 是做什么的

开发类 Skill 面向写代码、改代码和排查问题,一般会读写仓库文件,部分还会执行构建或测试命令。

dev-terminal 维护跨脚本执行的持久终端会话,支持 TUI 应用运行与屏幕输出捕获。它可在后台启动 server 进程,发送按键序列并调试终端程序,适用于 CLI 工具开发与自动化测试。可选 headed 模式查看实时画面,便于手动干预或问题排查,增强交互透明度。

使用前请安装 npm 包并启动 server,注意端口占用与防火墙设置,避免连接失败。适用宿主包括 Codex、Claude、Cursor、Gemini CLI,接入前应确认版本、权限和运行环境要求。

来源仓库内路径:skills/dev-terminal

上手步骤

4 步
  1. 1

    确认自己的宿主工具

    公开资料标注支持 Codex、Claude、Cursor、Gemini CLI;先确认自己用的是其中之一,再按对应工具的 Skills 目录放置。

  2. 2

    拿到安装命令

    可以直接复制的命令是 npx skills add https://github.com/parkerhancock/dev-terminal --skill dev-termina…;命令里的占位符换成自己的路径或密钥再执行。

  3. 3

    安装后重启宿主

    命令执行完要完全退出宿主工具再打开,让新装的 Skill 被重新扫描到,否则列表里可能看不到。

  4. 4

    按最小权限试用

    它需要接触敏感数据,先在测试目录或测试账号里跑,确认没问题再放到正式项目。

常见问题

6 条

dev-terminal 怎么安装?

在终端执行 npx skills add https://github.com/parkerhancock/dev-terminal --skill dev-terminal。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

dev-terminal 支持哪些 AI 工具?

公开资料标注支持 Codex、Claude、Cursor、Gemini CLI。其他工具只要能读取本地 Skill 目录,一般也能手动放进去,但要自己确认触发方式是否一致。

dev-terminal 需要什么权限?

权限级别是接触敏感数据。该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

dev-terminal 是免费的吗?

来源标注的许可证是 MIT。本站不代收任何费用,也不托管安装包;能否商用、要不要署名,以许可证原文为准。

dev-terminal 用的人多吗?

公开数据是累计安装 1,018 次,近一周安装 42 次,下载 333 次,GitHub 1 星。这些数字来自第三方目录统计,只能当作热度参考,不代表质量或安全性。

dev-terminal 安全吗?

第三方扫描有未通过项,风险较高。除非你确认过源码,否则不建议直接安装到常用环境。

值得选的地方和要先想清楚的地方

值得选的地方

  • 同时适配 4 种工具,换客户端不用重新找替代品。
  • 3 个目录同时收录,说明这个 Skill 流传比较广。
  • 许可证明确为 MIT,商用和分发边界好判断。

先想清楚的地方

  • 需要接触敏感数据,装到常用环境前要先评估影响面。
  • 第三方扫描存在提醒项或未通过项,需要人工逐条确认。
  • 第三方目录整理,版本更新和行为细节以来源仓库为准。

适不适合你

更适合

  • 已经在用 Codex,想直接扩展它能力的人。
  • 需要处理开发类任务、又不想每次都手写提示词的人。

不太适合

  • 环境里有大量密钥、凭证,不方便给第三方 Skill 读取的人。
  • 需要厂商级技术支持和 SLA 的生产系统。

开发类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
skill-vetter

技能审查员 · OpenClaw

8,779,018986操作浏览器MIT-0通过
ontology

本体论 · OpenClaw

6,636,960576只读MIT-0通过
entra-app-registration

内部应用程序注册 · Codex、Claude

6,631,872764接触敏感数据未明确通过
entra-app-registration

内部应用程序注册 · Codex、Claude

3,757,680201接触敏感数据未明确通过
nano-banana-pro

纳米香蕉专业版 · OpenClaw

3,526,898364接触敏感数据MIT-0有提醒
dev-terminal

开发终端 · 本页

1,0181接触敏感数据MIT未通过

按安装量取同分类前几名;数字来自第三方目录统计,只作热度参考,不代表质量结论。

SKILL.md 原文

Dev Terminal Skill

Terminal automation that maintains PTY sessions across script executions. Run TUI applications, send keystrokes, capture screen output, and debug terminal apps - all with persistent state.

Setup

Start the server in a background terminal:

cd dev-terminal && npm install && ./server.sh &

Wait for the Ready message before running scripts.

Headed Mode (Optional)

For visual debugging, start with browser UI:

cd dev-terminal && ./server.sh --headed &

This opens a browser window showing all terminals in real-time. Useful for:

  • Watching AI actions as they happen
  • Manual intervention if needed
  • Debugging TUI interactions

Writing Scripts

Run scripts inline using heredocs from the dev-terminal/ directory:

cd dev-terminal && npx tsx <<'EOF'
import { connect, sleep } from "./src/client.js";

const client = await connect();

// Create or get a named terminal
const term = await client.terminal("my-app", {
  command: "python",
  args: ["-m", "my_module"],
  cols: 120,
  rows: 40,
});

// Wait for app to start
await sleep(1000);

// Get screen snapshot
const snap = await term.snapshot();
console.log("=== SCREEN ===");
console.log(snap.text);

client.disconnect();
EOF

Shell Defaults

By default, terminals use:

  • Shell: User's default shell ($SHELL env var, e.g., zsh on macOS, bash on Linux)
  • Mode: Login shell (-l flag) - loads full profile (~/.zprofile, ~/.bash_profile)

This means your shell aliases, PATH, and environment are available.

Override the shell:

// Use a specific shell
const term = await client.terminal("my-term", {
  command: "bash",
  args: ["-l"], // keep as login shell
});

// Non-login shell (only loads ~/.bashrc, not ~/.bash_profile)
const term = await client.terminal("my-term", {
  command: "bash",
  args: [], // no -l flag
});

// Run a command directly (not a shell)
const term = await client.terminal("my-term", {
  command: "python",
  args: ["-m", "my_app"],
});

SSH Remote Terminals

Connect to remote servers via SSH. The API is identical to local terminals.

import { connect } from "./src/client.js";
import * as fs from "fs";
import * as os from "os";
import * as path from "path";

const client = await connect();

// SSH with private key
const term = await client.terminal("remote-server", {
  ssh: {
    host: "192.168.1.100",
    username: "deploy",
    privateKey: fs.readFileSync(path.join(os.homedir(), ".ssh/id_rsa"), "utf8"),
  },
});

// SSH with password
const term = await client.terminal("remote-server", {
  ssh: {
    host: "example.com",
    username: "admin",
    password: "secret",
  },
});

// SSH with agent (uses SSH_AUTH_SOCK)
const term = await client.terminal("remote-server", {
  ssh: {
    host: "example.com",
    username: "admin",
    agent: process.env.SSH_AUTH_SOCK,
  },
});

// SSH with custom port and encrypted key
const term = await client.terminal("remote-server", {
  ssh: {
    host: "example.com",
    port: 2222,
    username: "admin",
    privateKey: fs.readFileSync("/path/to/key", "utf8"),
    passphrase: "key-passphrase",
  },
});

SSH Options:

OptionTypeDescription
hoststringRemote hostname or IP (required)
portnumberSSH port (default: 22)
usernamestringSSH username (required)
passwordstringPassword authentication
privateKeystringPrivate key content (not path)
passphrasestringPassphrase for encrypted keys
agentstringPath to SSH agent socket

Notes:

  • SSH terminals don't have a pid (it's undefined)
  • All Terminal methods work the same (write, key, snapshot, etc.)
  • The remote shell is determined by the server, not local settings

Key Principles

  1. Small scripts: Each script does ONE thing (start app, check screen, send key)
  2. Observe output: Always check snapshot() to see current state
  3. Descriptive names: Use "claude-monitor", "vim-edit", not "term1"
  4. Terminals persist: disconnect() leaves terminals running for next script

Workflow Loop

  1. Write a script to perform one action
  2. Run it and observe the screen output
  3. Evaluate - what's displayed? Did it work?
  4. Decide - send more input or task complete?
  5. Repeat until done

Client API

const client = await connect();

// Get or create named terminal (uses default shell as login shell)
const term = await client.terminal("name");

// With options
const term = await client.terminal("name", {
  command: "python", // override shell/command
  args: ["-m", "my_app"], // override args (clears default -l flag)
  cols: 120,
  rows: 40,
  cwd: "/path/to/dir",
  env: { MY_VAR: "value" },
});

// List all terminal names
const names = await client.list();

// Close/kill a terminal
await client.close("name");

// Disconnect (terminals persist)
client.disconnect();

Terminal Methods

// Send raw input
await term.write("hello");

// Send special keys
await term.key("enter");
await term.key("up");
await term.key("ctrl+c");

// Send a line (adds Enter)
await term.writeLine("ls -la");

// Get screen state
const snap = await term.snapshot();
console.log(snap.text); // Plain text (no ANSI codes)
console.log(snap.lines); // Array of lines
console.log(snap.alive); // Process still running?

// Get SVG rendering (for visual analysis)
const svgSnap = await term.snapshot({ format: "svg" });
console.log(svgSnap.svg); // SVG string

// Resize
await term.resize(80, 24);

// Clear buffer
await term.clear();

// Wait for text to appear
const found = await term.waitForText("Ready", { timeout: 5000 });

// Wait for process to exit
const exitCode = await term.waitForExit({ timeout: 10000 });

Special Keys

Use term.key() with these names:

CategoryKeys
Arrowsup, down, left, right
Controlenter, tab, escape, backspace, delete
Ctrl+Xctrl+c, ctrl+d, ctrl+z, ctrl+l, ctrl+a, ctrl+e, ctrl+k, ctrl+u, ctrl+w, ctrl+r
Functionf1 - f12
Navigationhome, end, pageup, pagedown, insert

Example: Debug a TUI App

cd dev-terminal && npx tsx <<'EOF'
import { connect, sleep } from "./src/client.js";

const client = await connect();

// Start the TUI
const term = await client.terminal("claude-monitor", {
  command: "../.venv/bin/python",
  args: ["-m", "claude_monitor"],
  cols: 120,
  rows: 40,
  cwd: "..",
});

// Wait for it to render
await sleep(2000);

// Capture screen
const snap = await term.snapshot();
console.log("=== SCREEN OUTPUT ===");
console.log(snap.text);
console.log("=== ALIVE:", snap.alive, "===");

client.disconnect();
EOF

Example: Interactive Session

# Script 1: Start app
cd dev-terminal && npx tsx <<'EOF'
import { connect, sleep } from "./src/client.js";
const client = await connect();
const term = await client.terminal("my-tui", {
  command: "htop",
});
await sleep(1000);
const snap = await term.snapshot();
console.log(snap.text);
client.disconnect();
EOF

# Script 2: Send keys (terminal persists!)
cd dev-terminal && npx tsx <<'EOF'
import { connect, sleep } from "./src/client.js";
const client = await connect();
const term = await client.terminal("my-tui"); // Reconnect to existing
await term.key("down");
await term.key("down");
await sleep(500);
const snap = await term.snapshot();
console.log(snap.text);
client.disconnect();
EOF

# Script 3: Quit
cd dev-terminal && npx tsx <<'EOF'
import { connect } from "./src/client.js";
const client = await connect();
const term = await client.terminal("my-tui");
await term.write("q");
client.disconnect();
EOF

Error Recovery

If something goes wrong, check the terminal state:

cd dev-terminal && npx tsx <<'EOF'
import { connect } from "./src/client.js";

const client = await connect();

// List all terminals
const terminals = await client.list();
console.log("Active terminals:", terminals);

// Check specific terminal
if (terminals.includes("my-app")) {
  const term = await client.terminal("my-app");
  const snap = await term.snapshot();
  console.log("Alive:", snap.alive);
  console.log("Exit code:", snap.exitCode);
  console.log("Last output:", snap.lines.slice(-20).join("\n"));
}

client.disconnect();
EOF

Tips

  • TUI apps need time: Use sleep() after starting to let them render
  • Check alive status: TUI might crash - check snap.alive
  • Clear for fresh state: Use term.clear() before important snapshots
  • Large output: snap.lines gives the last ~120 lines (3x terminal height)

平台分布

Claude Code

28.23%
按下载量换算94

OpenCode

23.02%
按下载量换算77

trae

19.32%
按下载量换算64

Antigravity

12.06%
按下载量换算40

windsurf

9.09%
按下载量换算30

Codex

3.3%
按下载量换算11

安全审计

Gen Agent Trust Hub

未通过

Socket

未通过

Snyk

未通过

权限和风险

接触敏感数据

该 Skill 可能接触密钥、Token、环境变量或敏感配置,应进入高风险复核队列,默认不自动发布。

安装前确认

本站仅展示第三方公开信息,不托管安装包,不提供自动安装或运行环境。安装前应自行审查源码、依赖和命令行为。来源安全扫描存在提醒或未通过项,安装前请自行复核。

继续浏览同类 Skills