Token导航 LogoToken导航

Agent Skill · 研究检索

skill-extractor(技能提取器)

skill-extractor 用于查找、检索和筛选相关信息,适合在 OpenClaw 中需要根据关键词、任务场景或来源线索快速定位候选结果时使用。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。

skill-extractor 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为需要联网;许可证为 MIT-0,第三方扫描有提醒项,按安装量在研究检索类的 73,675 个 Skill 里排第 26,627。

总安装 3,528周安装 150GitHub Stars 公开资料未说明下载 1,236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-0
版本
1.8.1

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:skill-extractor(技能提取器)
来源仓库:https://github.com/seph1709/skill-extractor
安装命令:
openclaw skills install skill-extractor
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。该命令会通过 OpenClaw 从第三方来源获取 Skill;本站只展示命令,不托管安装包,也不自动执行。

ClawHubOpenClaw
openclaw skills install skill-extractor

这个 Skill 是做什么的

研究检索类 Skill 主要解决「找资料、核对事实、把来源整理清楚」的问题,通常会让助手先检索再总结。

将已安装的 OpenClaw 技能打包为可共享的 ZIP 文件。适合导出技能结构、生成 STRUCTURE.md 供他人复用或审计。自动检测外部依赖并暂存运行时文件,提升技能移植性。执行前应确认目标路径可写,避免覆盖重要数据。建议在非生产环境测试导出结果完整性。

skill-extractor 属于研究检索类 Skill,可作为该场景下的辅助能力补充。

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 OpenClaw;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 openclaw skills install skill-extractor;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

    它需要需要联网,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

skill-extractor 怎么安装?

在终端执行 openclaw skills install skill-extractor。装完完全退出宿主工具再打开,让它重新扫描本地 Skill。

skill-extractor 支持哪些 AI 工具?

公开资料标注支持 OpenClaw。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。

skill-extractor 需要什么权限?

权限级别是需要联网。该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

skill-extractor 是免费的吗?

来源标注的许可证是 MIT-0。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。

skill-extractor 用的人多吗?

公开数据是累计安装 3,528 次,近一周安装 150 次,下载 1,236 次。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。

skill-extractor 安全吗?

第三方扫描有提醒项,说明存在需要人工确认的行为。安装前请打开源码看清它读了什么、联网到哪里。

值得选的地方和要先想清楚的地方

值得选的地方

  • 专注 OpenClaw 一种宿主,触发方式和使用习惯不用来回切换。
  • 许可证明确为 MIT-0,商用和分发边界好判断。

先想清楚的地方

  • 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 OpenClaw,想直接扩展它能力的人。
  • 需要处理研究检索类任务、又不想每次都手写提示词的人。

不太适合

  • 需要厂商级技术支持和 SLA 的生产系统。

研究检索类 Skill 对比

6 个同类
Skill总安装GitHub 星权限许可证扫描
find-skills

寻找技能 · Codex、Claude

31,199,81316,486操作浏览器MIT有提醒
self-improving-agent

自我改进 Agent · OpenClaw

16,464,0023,396接触敏感数据MIT-0通过
soultrace

灵魂追踪 · Codex、Claude

7,588,94412需要联网未明确有提醒
self-improving

自我完善 · OpenClaw

6,829,2841,046需要联网MIT-0通过
microsoft-foundry

微软代工厂 · Codex、Claude

6,614,144764调用外部服务未明确有提醒
skill-extractor

技能提取器 · 本页

3,528—需要联网MIT-0有提醒

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

name
skill-extractor
description
Export any installed OpenClaw skill into a shareable ZIP: detects & stages external runtime files, generates STRUCTURE.md for LLM-guided install. Reads and packages local files only — no network calls, no APIs, no external transmissions.
metadata
{"openclaw":{"emoji":"📦"}}

skill-extractor

Package any installed OpenClaw skill into a clean, shareable ZIP. External runtime files referenced in SKILL.md are detected, staged under _external/, and documented — so a new install knows exactly where every file belongs and can reproduce full functionality.

All file operations are local only. Nothing is transmitted anywhere. The user confirms what gets included before anything is zipped.

Agent Rules

  • Always list available skills before asking for selection (unless skill name is already given)
  • Always work on a staging copy — never modify the original skill directory or any external paths
  • Always show the user what external files were found and get explicit confirmation before zipping
  • Files are packaged as-is — values are not modified. Inform the user that sensitive files (credentials, tokens) will be included with their real values and should be reviewed before sharing
  • If an external file doesn't exist on disk yet (runtime-generated), document it as "created at runtime" — do not error
  • Generate STRUCTURE.md inside the staging folder before zipping
  • Default ZIP output: the user's Desktop — confirm with user first
  • If ZIP already exists at target, overwrite it
  • Clean up staging after a successful ZIP
  • If any step fails, leave staging intact and report clearly

Step 1 — List Available Skills

Scan all known OpenClaw skill locations (workspace skills folder, user-local skills folder, and the bundled npm package skills folder) for subdirectories that contain a SKILL.md. Present the list with skill name and source. Ask which to export.


Step 2 — Locate Skill

Find the skill folder by name. If not found, report and stop.


Step 3 — Stage Skill Files

Create a hidden temp staging folder inside the workspace named after the skill. Copy all files from the skill directory into it. Originals are never touched.


Step 4 — Detect External Files

Read the SKILL.md from the original skill directory. Extract all path-like strings that begin with a user home or app-data prefix (home dir shorthands and platform app-data equivalents). Resolve each to an absolute path.

For each path:

  • If it is a file that exists: add it to the external files list.
  • If it is a directory that exists: add all files inside it recursively to the list.
  • If it doesn't exist yet: record it as a "created at runtime" entry.

Do not copy anything yet. Build the list first — it goes to the user for review in the next step.


Step 5 — Confirm with User

Present the user with everything that will be included in the ZIP:

  1. All files from the skill directory
  2. The full list of detected external files with their resolved paths

Clearly warn: "These files will be packaged with their real values — including any credentials, tokens, or sensitive config. Review before sharing the ZIP."

Ask: *"Proceed with packaging these files?"*

  • If yes: stage the external files into _external/, mirroring the directory structure relative to the user's home, then continue.
  • If no: abort and clean up staging. Do not create a ZIP.

Step 6 — Generate STRUCTURE.md

Write STRUCTURE.md into the staging dir with these sections in order:

Header — skill name, generation timestamp, one-line purpose statement.

Folder Layout — ASCII tree of the staging dir. Directories before files at each level, both sorted alphabetically. Use standard tree connectors (├──, └──, │). Directory names end with /.

File Descriptions table — two columns: relative file path and one-line purpose. Describe each file by what it does rather than its type. Well-known skill files have fixed descriptions:

FileDescription
SKILL.mdMain skill instructions. The LLM agent reads this to understand purpose, setup, and usage.
_meta.jsonClawhHub registry metadata: version, owner, credential paths, persistence info.
STRUCTURE.mdThis file. Auto-generated folder map and install guide.
Any other fileInfer purpose from filename, location, and content — describe what it does in plain English.

External Files table — only if external files were detected. Three columns: file path inside the ZIP | target install path on the machine | notes. Determine the notes value by this logic:

ConditionNotes
Path suggests credentials or configPackaged with real values — review before sharing.
Path suggests a worker or background scriptExtracted from SKILL.md at runtime — included here for reference.
Path suggests a log, pid, or state fileRuntime-generated. Recreated automatically on first run.
File did not exist at export timeNot present at export — created automatically when the skill runs.
Anything elseExternal runtime file. Review SKILL.md for usage.

Follow the table with a brief note on how to place _external/ files at their target paths on both Windows and Unix systems.

Install Instructions — three options:

  • Option A: install via ClawhHub if the skill is published
  • Option B: manual — copy the skill folder into the workspace skills directory, place any _external/ files at their target paths, confirm with openclaw skills list
  • Option C: local clawhub install, then handle _external/ files manually

Footer — credit line linking to the skill on ClawhHub.


Step 7 — Zip and Deliver

Confirm the output path with the user (default: the user's Desktop). Remove any existing ZIP at that path first. Compress the staging dir. Report the saved path and file size. Delete the staging folder.


Error Reference

ProblemCauseFix
Skill not foundName mismatchCheck spelling; run openclaw skills list
Access denied on copyFile ownership issueRun as admin or check permissions
ZIP creation failsDisk full or missing compression supportFree space or update runtime
Staging not cleanedZIP step failedDelete the staging folder inside the workspace manually
External file missingRuntime-generated, not yet createdSafe to skip — document as "created at runtime"
User declined confirmationSensitive files flagged by userAbort — no ZIP created, staging cleaned up

平台分布

OpenClaw

94%
按下载量换算1,162

安全审计

VirusTotal

可疑

ClawScan

通过

Static analysis

通过

权限和风险

需要联网

该 Skill 可能需要联网访问来源站点、仓库或外部 API;具体网络访问范围需要结合源码和 README 复核。

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。目前只有一个公开来源,建议同时核对源仓库。

继续浏览同类 Skills