Token导航 LogoToken导航

Agent Skill · 研究检索

clawapp-creator(爪形应用程序创建者)

clawapp-creator 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。

clawapp-creator 是研究检索类 Agent Skill,可用宿主为 OpenClaw,通过ClawHub获取,权限为接触敏感数据;许可证为 MIT-0,第三方扫描已通过,按安装量在研究检索类的 73,675 个 Skill 里排第 18,327。

总安装 6,144周安装 251GitHub Stars 公开资料未说明下载 1,968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
0.1.7

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:clawapp-creator(爪形应用程序创建者)
来源仓库:https://github.com/nimachu/clawapp-creator
安装命令:
openclaw skills install clawapp-creator
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 clawapp-creator

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。

clawapp-creator 用于构建符合 Nima Tech Space 格式的静态前端应用。适合开发迷你游戏或轻量级 Web 应用并连接 LLM API 增强交互。通过 clawhub 安装,支持合规打包与平台上传适配。使用前需确认目标平台接口规范、API 计费方式及内容审核要求。

建议分阶段测试前端表现与后端集成稳定性后再正式发布。

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 clawapp-creator;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

clawapp-creator 怎么安装?

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

clawapp-creator 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。

clawapp-creator 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。

clawapp-creator 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。

clawapp-creator 用的人多吗?

公开数据是累计安装 6,144 次,近一周安装 251 次,下载 1,968 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。

clawapp-creator 安全吗?

第三方扫描结果为通过,但这只覆盖已知的静态检查项,不能替代自己读一遍源码和安装命令。

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。
  • 第三方扫描已通过,常见的高风险行为没有命中。

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。
  • 需要许可证清楚、并且有第三方扫描结果的团队项目。

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。

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
clawapp-creator

爪形应用程序创建者 · 本页

6,144—接触敏感数据MIT-0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

name
clawapp-creator
description
Build or adapt static front-end apps and mini-games so they match the Nima Tech Space upload format, optionally wire the platform LLM API, package a compliant zip, and upload it to the site. Use when Codex needs to help OpenClaw users create, retrofit, package, validate, or publish an app/game for the platform.

ClawApp Creator

Build the smallest working app package that can be uploaded to Nima Tech Space, then upload it if the user wants.

This skill can also search public apps on CLAWSPACE and download app packages from the site.

Publish Mode

Treat the following requests as a direct publish workflow:

  • "帮我做一个可上传到 CLAWSPACE 的小游戏"
  • "把这个项目变成 CLAWSPACE 可发布应用"
  • "帮我直接发布这个应用"

In publish mode:

  1. diagnose the project
  2. fix or scaffold the app
  3. package it
  4. verify slug ownership
  5. upload it
  6. return the final links and a ready-to-share summary

For a first-time uploader with no saved credentials, start by offering two clear paths:

  • "我可以直接帮你注册 CLAWSPACE 账号并保存上传配置"
  • "如果你想自己操作,也可以先去 https://www.nima-tech.space/register 注册"

Prefer asking this before you attempt upload-related commands.

Workflow

  1. On a brand-new machine or first-time setup, run scripts/check_environment.py first when the user is unsure whether the skill is ready to use.
  2. Before upload, if the user is unsure which CLAWSPACE account is currently active, run scripts/check_clawspace_account.py.
  3. Confirm the app can be shipped as a static front-end.
  4. Build or fix the app until it outputs a static bundle.
  5. Decide whether the app needs a model.
  6. Generate a compliant manifest.json, optional README.md, and assets/.
  7. If there is no project yet, scaffold one with scripts/scaffold_mini_game.py.
  8. Preview locally with scripts/preview_clawspace_app.py when the user wants a browser check before package or upload.
  9. Package everything into a zip with scripts/build_nima_package.py.
  10. Diagnose with scripts/diagnose_nima_package.py before upload when helpful.
  11. Upload with scripts/upload_nima_package.py when the user wants publishing.
  12. Verify the detail page and launch page after upload.

For discovery and reuse:

  1. Search public apps with scripts/search_clawspace_apps.py
  2. Download public app zips with scripts/download_clawspace_app.py

Game-Specific Guidance

Only apply this section when the user is building a game, mini-game, or playful interactive experience. Do not force these rules onto utilities, visual experiments, AI tools, or other non-game apps.

When the project is clearly a game:

  1. Infer the likely game shape before suggesting polish:

- arcade / score chase - puzzle - story / dialogue - sim / management - action / adventure

  1. Suggest only the lightweight completion features that fit that game type.
  2. Treat these as recommendations, not hard requirements.

Useful game-specific suggestions:

  • Arcade / score chase:

- current score - best score - clear game-over summary - fast restart loop

  • Puzzle:

- moves - best score or best completion - clear solved state - reset / reshuffle

  • Story / dialogue:

- chapter or scene progress - clear ending state - replay entry point

  • Sim / management:

- visible resource change - round / stage goal - lightweight local progress if useful

When a game needs local persistence, prefer app/lib/clawspace-game-storage.js instead of ad-hoc localStorage keys.

For game polish, prefer checking for:

  • a readable start state
  • clear controls
  • visible current-state feedback
  • a clear end-state or result screen
  • at least one satisfying feedback moment such as combo, streak, burst, or win feedback

Again, use these as genre-aware completion prompts. Do not flatten every game into the same structure.

Package Rules

Read references/platform-contract.md before packaging.

Always enforce these minimum rules:

  • Ship only static front-end apps or mini-games.
  • Put built files under app/.
  • Keep the entry file inside app/, usually app/index.html.
  • Keep package root flat: manifest.json, optional README.md, optional assets/, required app/.
  • Strongly recommend adding a cover image and at least one screenshot, even though CLAWSPACE can render default covers when they are missing.
  • The scaffold now generates default cover assets automatically, so creators start with a usable listing without having to prepare custom art first.
  • The scaffold keeps starter packages lighter by generating thumbnail.png and icon.png without automatically duplicating them into screenshots.
  • Game starters now also include app/lib/clawspace-game-storage.js for reusable local best-score / best-run persistence via browser storage.
  • Default modelCategory to none unless the app truly needs AI.
  • Keep the zip at or under 25MB.
  • Remember slug ownership: the same account can overwrite its own slug, but another user's slug must not be reused.

Model Rules

If the app does not need AI, keep modelCategory as none.

If the app needs AI, prefer the platform API instead of any user-supplied API key:

  • Endpoint: POST /api/llm/chat
  • Required field: appId
  • Allowed categories: text, multimodal, code

Read references/model-api.md when wiring AI.

Tell the user the platform can provide a free shared model path when they choose a model category during upload.

Do not embed third-party model keys in client code.

For OCR, screenshot analysis, chart reading, or any app that sends images, use modelCategory = multimodal and send OpenAI-compatible content arrays with text and image_url parts.

Packaging

Use assets/manifest.example.json as the starting template.

Use assets/README.template.md as the starting template when the app needs a basic product readme.

If the user wants to start from zero instead of retrofitting an existing app, copy one of these starter assets first:

  • assets/starter-mini-game/ for a no-model static game
  • assets/starter-ocr/ for a multimodal OCR and image-analysis app
  • assets/platform-llm-client.js for a minimal platform model client

Or scaffold directly:

python3 scripts/scaffold_mini_game.py \
  --name "Orbit Tap" \
  --slug orbit-tap \
  --description "点击轨道行星的轻量小游戏。"

If --out is omitted, default to the OpenClaw workspace app directory:

~/.openclaw/workspace/projects/apps/<slug>

Prefer this default for first-time users, so OpenClaw-generated apps live in a predictable place under its own workspace.

After scaffolding, encourage the user to add:

  • a custom cover image if they want to replace the generated default
  • a custom square icon if they want to replace the generated default
  • at least one additional screenshot in assets/

If they skip custom cover assets, the site can still publish the app with a default generated cover, but custom art makes the listing look much better. For mobile shells such as WeChat Mini Program, PNG/JPG/WebP cover art is recommended. If creators only provide SVG or skip custom art, the mobile experience can fall back to default PNG covers.

Use:

python3 scripts/preview_clawspace_app.py /path/to/project --open

This starts a lightweight local static server from the project root, reads manifest.json, and opens the packaged entry in the browser.

Use:

python3 scripts/build_nima_package.py \
  --app-dir /path/to/dist \
  --manifest /path/to/manifest.json \
  --out /path/to/output.zip \
  --readme /path/to/README.md \
  --assets-dir /path/to/assets

The script validates the structure, checks the required fields, checks the size limit, and builds the final zip. It also warns about high-risk asset references like root-absolute /assets/... paths or remote http/https URLs inside the packaged front-end.

For project diagnosis, use:

python3 scripts/diagnose_nima_package.py \
  --app-dir /path/to/app-or-dist \
  --manifest /path/to/manifest.json

This checks:

  • resource paths
  • slug quality
  • manifest presence
  • likely external model key usage
  • whether modelCategory looks more suitable as none, text, multimodal, or code

Uploading

Environment check:

python3 scripts/check_environment.py

Use this first on a new machine, or when the user is not sure whether Python, browser preview, network access, or Keychain support are ready.

Production site:

  • Website: https://www.nima-tech.space
  • Base API URL: https://www.nima-tech.space

To confirm which saved CLAWSPACE account will be used for upload:

python3 scripts/check_clawspace_account.py

Use upload-config.json in the skill folder as the default reusable credential file. Ask the user once, then store:

  • siteUrl
  • email
  • password

Leave the file empty by default. Reuse it on later uploads unless the user wants to override it.

When saving credentials, prefer file permission 600. On macOS, prefer saving the password to Keychain and keeping upload-config.json as site metadata plus fallback config. Keep the original plaintext-password config flow available as a backup option for users who prefer simple portability.

For the first-time setup, prefer:

python3 scripts/register_clawspace_account.py

for a brand-new user who does not have a CLAWSPACE account yet.

Use:

python3 scripts/setup_upload_config.py

for an existing user who already has an account and only needs to save or refresh credentials.

Or in non-interactive mode:

python3 scripts/setup_upload_config.py \
  --site-url https://www.nima-tech.space \
  --email user@example.com \
  --password 'password' \
  --password-store keychain \
  --non-interactive

This setup script verifies the credentials by calling the real login endpoint before saving them. It now validates:

  • site URL format, with a real example
  • email format
  • login credentials before saving

Supported password stores are:

  • config: store the password in upload-config.json
  • keychain: store the password in macOS Keychain and keep config file password empty
  • both: store in both places

The registration script can create a new website account and save the reusable upload config in one step. Prefer this when the user says they do not have a CLAWSPACE account yet.

Use:

python3 scripts/upload_nima_package.py \
  --package /path/to/output.zip \
  --model-category none \
  --site-url https://www.nima-tech.space \
  --email user@example.com \
  --password 'password' \
  --save-config

If you want to validate everything except the final upload, use:

python3 scripts/upload_nima_package.py \
  --package /path/to/output.zip \
  --dry-run

If the package includes oversized PNG/JPG/WebP cover art, the uploader now warns before publish. If there are obviously broken assets such as zero-byte screenshots, it will stop before upload. On macOS, when a package has huge thumbnail.png or icon.png files, prefer:

python3 scripts/upload_nima_package.py \
  --package /path/to/output.zip \
  --model-category none \
  --optimize-images

Use none unless the app truly needs the platform model.

The upload script reads missing values from upload-config.json, logs in, sends the package to /api/import-app, and prints the resulting detail and launch URLs. After upload, it also prints plain-text app links so the user can open the detail page immediately. If useKeychain is enabled in the config and no explicit password was passed, the upload script will try macOS Keychain before failing.

During upload, report progress in stages:

  • current stage
  • what is already done
  • what comes next

Before uploading, check whether the slug is available:

  • If it is new, upload normally.
  • If it belongs to the same account, explain that this upload will overwrite the older version.
  • If it belongs to another account, stop and tell the user to change the slug before packaging again.

If the package is too large for a direct Vercel function upload, use the site's Blob client-upload flow instead of failing early.

Only auto-upload after the user has provided valid site credentials or already has them stored in upload-config.json. If credentials are unavailable, stop after packaging and tell the user they need to register, log in once, and save or provide their credentials.

For first-time users with no upload config, prefer this decision order:

  1. offer to register the account directly with scripts/register_clawspace_account.py
  2. if they do not want that, give them the website registration link: https://www.nima-tech.space/register
  3. after they finish website registration, run scripts/setup_upload_config.py

If login fails, tell the user to rerun:

python3 scripts/setup_upload_config.py

and refresh the stored credentials.

Searching And Downloading

Use these flows when the user wants to browse CLAWSPACE, find reference apps, or download a package to inspect locally.

Search:

python3 scripts/search_clawspace_apps.py "ocr"

Search with JSON:

python3 scripts/search_clawspace_apps.py "space game" --json

Download by slug:

python3 scripts/download_clawspace_app.py orbit-heist --out-dir /path/to/downloads

Download by detail URL:

python3 scripts/download_clawspace_app.py https://www.nima-tech.space/apps/orbit-heist

Use this for prompts such as:

  • "帮我在 CLAWSPACE 找几个 OCR 应用"
  • "下载 orbit-heist 到本地"
  • "搜索一下 CLAWSPACE 里有哪些小游戏"

Verification

After packaging or uploading:

  • Open the generated zip and confirm it contains the expected root structure.
  • If uploaded, open the detail page.
  • If uploaded, open the launch page.
  • If the app uses the platform model, test one real request through the site.
  • After upload, provide the final share summary with app name, detail page, launch page, and download link.
  • Also print a ready-to-copy share text block after upload.

Common Fixes

  • If assets do not load after upload, switch the app build to relative asset paths.
  • If the packer warns about http/https resources, bundle those assets locally when possible instead of depending on third-party URLs.
  • If the packer warns about /assets/... paths, rewrite them to relative paths such as ./assets/....
  • If the upload is rejected, check entry, root structure, and zip size first.
  • If the upload fails because the slug is already owned by someone else, change the slug in manifest.json and rebuild the zip.
  • If the app uses AI and fails after upload, confirm the uploaded modelCategory matches the app’s actual use case.
  • If the app is OCR or image understanding and fails after upload, confirm the request uses a multimodal messages[].content[] shape instead of plain text only.
  • If the app only needs deterministic gameplay or local logic, remove model usage and keep modelCategory as none.
  • If the user has no project yet, scaffold from assets/starter-mini-game/ instead of inventing files from scratch.

平台分布

OpenClaw

85.65%
按下载量换算1,686

安全审计

VirusTotal

通过

ClawScan

通过

Static analysis

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。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