📦 Builtin Tools — 技能工具

v1.0.0

跨平台基础工具集 — 16 个独立可组合的 Python 脚本，替代 Agent 平台缺失的基础工具能力。文件系统（浏览/搜索/读写/替换/删除）、内容搜索（正则）、网络（搜索/抓取/预览）、运行时安装、持久化记忆、定时任务、任务管理。纯 Python 标准库，零外部依赖，跨 Windows/macOS/L...

0· 50·0 当前·0 累计

by @wangjiaocheng (波动几何)·MIT-0

存储部署系统工具设计工具

下载技能包

License

MIT-0

最后更新

2026/4/14

安全扫描

VirusTotal

可疑

查看报告

OpenClaw

安全

medium confidence

The skill's code and runtime instructions are internally consistent with a multi-tool 'builtin tools' purpose: it provides file, search, web fetch, command execution, installation and simple persistence capabilities and does not request unrelated credentials or external install sources beyond run-time downloads from official project domains.

评估建议

This package is internally coherent for a local ’builtin tools’ toolkit, but it grants powerful local privileges: it can read/write/delete files, run arbitrary shell commands, fetch arbitrary URLs, and download/extract binaries. Only install or enable it if you trust the author and the code. Consider: run it in a sandbox or isolated account first; disable autonomous invocation if your platform allows it (so an agent cannot call 'command' mode without your approval); review the remaining unshown ...

详细分析 ▾

✓ 用途与能力

Name/description advertise cross-platform small utilities (filesystem, search, web fetch, install, memory, tasks) and the package includes scripts that implement exactly those capabilities. Required env/config/credentials are none, which aligns with a pure-Python toolset.

ℹ 指令范围

SKILL.md and scripts allow broad operations: arbitrary shell command execution (execute_command mode 'command'), reading/writing/deleting files, recursive searches, web fetching, and runtime binary downloads. This is consistent with the advertised purpose, but these capabilities are high-privilege: the agent (or a caller) can read/write most files, execute arbitrary shell commands, and download/expand archives. The delete_file script blocks literal root and user home deletion, but most other filesystem paths are writable by the skill. execute_command uses a shell (sh -c or PowerShell -Command), so commands are interpreted by the platform shell.

✓ 安装机制

No install spec; skill is instruction + bundled Python scripts (no remote install). The only runtime downloader is install_binary.py which pulls from official python.org and nodejs.org URLs and extracts archives — expected for a runtime installer and documented in SKILL.md.

✓ 凭证需求

The skill declares no required environment variables or credentials. Scripts copy os.environ when spawning subprocesses and accept optional env overrides for executed commands, which is normal for a tool that runs local commands. There are no hard-coded secrets or requests for unrelated cloud credentials.

✓ 持久化与权限

always is false and the skill does not autonomously modify other skills or global agent config. It persists files under user home (e.g., ~/.builtin-tools, ~/.workbuddy memory dir) and can write todo/memory files; this matches the stated persistent memory feature and is scoped to the user's filesystem rather than system-wide privilege escalation.

安全有层次，运行前请审查代码。

License

MIT-0

可自由使用、修改和再分发，无需署名。

查看条款 ↗

运行时依赖

无特殊依赖

版本

latestv1.0.02026/4/13

Initial release of builtin-tools: a cross-platform Python toolkit of 16 independent scripts, replacing missing basic utilities in agent platforms. - Provides tools for file system operations, content search (including regex), networking (search/fetch/preview), runtime installation, persistent memory, scheduling, and task management. - All scripts use only the Python standard library (no external dependencies); runs on Windows, macOS, and Linux. - Unified JSON input/output protocol and error codes for consistency. - Features a self-bootstrap design: execute_command.py orchestrates all tools; only one Python command is needed for full functionality. - Includes strict security policies (e.g., limits on shell use and destructive operations).

● 可疑

安装命令

点击复制

官方npx clawhub@latest install builtin-tools

镜像加速npx clawhub@latest install builtin-tools --registry https://cn.longxiaskill.com

技能文档

# builtin-tools — 跨平台基础工具集

定位

当宿主 Agent 平台缺少某个基础工具时，用本技能的脚本补足。所有脚本纯 Python 标准库实现，零外部依赖，跨 Windows/macOS/Linux。

核心规则

优先用宿主平台内置工具 — 仅在平台不支持时回退到本技能脚本
通过 execute_command 调用 — 所有脚本接受 JSON 输入，输出 JSON 结果
自举设计 — execute_command.py 可调度所有其他脚本，只需平台支持执行一条 Python 命令

脚本清单

| 脚本 | 功能 | 调用方式 | |------|------|---------| | execute_command.py | 自举入口 + 命令执行 + 管道串联 | 见下方详细说明 | | list_dir.py | 目录浏览 | python list_dir.py '{"path":".","ignore":[".log"]}' | | search_file.py | 文件名搜索 | python search_file.py '{"pattern":".py","recursive":true}' | | read_file.py | 读取文件 | python read_file.py '{"path":"f.txt","offset":0,"limit":50}' | | write_file.py | 写入文件 | python write_file.py '{"path":"f.txt","content":"hello"}' | | replace_in_file.py | 精确替换 | python replace_in_file.py '{"path":"f.txt","old":"a","new":"b"}' | | delete_file.py | 删除文件 | python delete_file.py '{"path":"f.txt"}' | | search_content.py | 正则内容搜索 | python search_content.py '{"pattern":"TODO","path":"."}' | | web_search.py | 网页搜索 | python web_search.py '{"query":"Python","engine":"duckduckgo"}' | | web_fetch.py | 网页抓取 | python web_fetch.py '{"url":"https://..."}' | | preview_url.py | 浏览器预览 | python preview_url.py '{"url":"https://..."}' | | install_binary.py | 运行时安装 | python install_binary.py '{"type":"python","version":"3.13.12"}' | | update_memory.py | 持久化记忆 | python update_memory.py '{"action":"create","title":"...","content":"..."}' | | automation_update.py | 定时任务 | python automation_update.py '{"mode":"create","name":"...","prompt":"...","rrule":"..."}' | | todo_write.py | 任务管理 | python todo_write.py '{"todos":[{"id":"1","status":"in_progress","content":"..."}]}' |

execute_command.py 详细用法

模式 1：执行系统命令

json
{"mode": "command", "command": "ls -la", "timeout": 30, "cwd": "/tmp"}



模式 2：执行内置脚本

json
{"mode": "script", "script": "list_dir", "params": {"path": "."}}



模式 3：打开 URL/文件

json
{"mode": "open", "url": "https://example.com"}



模式 4：管道串联多个脚本

json
{"mode": "pipe", "chain": ["search_file", "search_content"], "params": {"pattern": ".py"}}

`快捷模式`bash python execute_command.py search_file '{"pattern":"

.py"}'
python execute_command.py --command "echo hello"



JSON 协议

所有脚本遵循统一协议：

输入：JSON 字符串（命令行第一个参数 或 stdin）

输出：{"status":"ok","data":{...},"message":"ok"} 或 {"status":"error","code":N,"message":"..."}


Exit Code：0=成功, 1=参数错误, 2=执行失败, 3=不支持的平台

`目录结构`


builtin-tools/
├── SKILL.md                    # 本文件
├── scripts/
│   ├── common.py               # 公共层（Shell执行/JSON协议/路径处理）
│   ├── execute_command.py      # 自举入口
│   ├── list_dir.py             # 目录浏览
│   ├── search_file.py          # 文件名搜索
│   ├── read_file.py            # 读取文件
│   ├── write_file.py           # 写入文件
│   ├── replace_in_file.py      # 精确替换
│   ├── delete_file.py          # 删除文件
│   ├── search_content.py       # 正则内容搜索
│   ├── web_search.py           # 网页搜索
│   ├── web_fetch.py            # 网页抓取
│   ├── preview_url.py          # 浏览器预览
│   ├── install_binary.py       # 运行时安装
│   ├── update_memory.py        # 持久化记忆
│   ├── automation_update.py    # 定时任务
│   └── todo_write.py           # 任务管理
└── references/
    └── protocol.md             # 接口协议详细文档



安全策略

execute_command.py 不使用 shell=True（Windows 用 list 参数传递给 powershell.exe）

delete_file.py 禁止删除根目录和用户主目录

replace_in_file.py 默认限制替换次数（防止误操作）

install_binary.py` 下载后校验文件完整性

数据来源：ClawHub ↗ · 中文优化：龙虾技能库