← 返回 Skills 市场
جاك العلم
作者
Abdullah AlRashoudi
· GitHub ↗
· v2.0.3
725
总下载
0
收藏
2
当前安装
5
版本数
在 OpenClaw 中安装
/install personal-shopper
功能描述
Personal Shopper — multi-agent product/service research and recommendation for Saudi Arabia. USE WHEN: - User asks to find, compare, recommend, or buy a prod...
安全使用建议
This skill appears internally consistent and lightweight: it does live web searches (including headless snapshots for JS-heavy pages) and writes an Arabic RTL HTML report to shopping-reports/. It asks for no credentials and has no installer. Things to consider before installing: 1) Reports embed external fonts and may link to retailer images and product pages — when you open a generated report your browser may make external requests. 2) The skill will follow links to third-party sellers; always verify seller reputation and URLs before clicking 'اشتري الآن' links. 3) The skill uses a headless/browser-fallback tool (camofox) for some pages — this is expected but means web content will be rendered/extracted; review network allowlist settings if you want to restrict which domains the skill may contact. 4) Autonomous invocation is allowed by default on the platform (not set to always:true here); if you want manual control, keep it user-invocable only. Overall: coherent for its purpose, but treat generated links and external resources with normal caution.
功能分析
Type: OpenClaw Skill
Name: personal-shopper
Version: 2.0.3
The OpenClaw AgentSkills skill bundle for 'Personal Shopper' is classified as benign. The skill is highly structured, detailing a multi-agent system for product research and HTML report generation. Crucially, it includes explicit security constraints: a strict domain allowlist for network access, limited file read/write access to specific directories (`shopping-reports/`, `references/`), and a clear instruction for the agent to treat all fetched web content as untrusted. There is no evidence of intentional malicious behavior such as data exfiltration, unauthorized execution, persistence mechanisms, or prompt injection attempts against the agent itself. The provided Python code for screenshot embedding also includes path traversal prevention, reinforcing a security-conscious design.
能力评估
Purpose & Capability
Name/description (Personal Shopper for Saudi Arabia) align with the declared behavior: multi-agent web research, price checks, coupons, and an HTML report. Tools referenced (web_search, web_fetch, camofox, sessions_spawn) are expected for scraping JS-heavy retailer pages and producing a report. No unrelated credentials, binaries, or config paths are requested.
Instruction Scope
SKILL.md confines actions to web searches/fetches, spawning sub-agents for structured research, and writing HTML to shopping-reports/. It explicitly limits file reads to references/ and shopping-reports/screenshots/. Minor note: the HTML template includes external Google Fonts and may reference retailer-hosted product images (causing external requests when the report is opened); this is expected for a report but is an external-network behavior to be aware of.
Install Mechanism
Instruction-only skill with no install spec and no code files — lowest install risk. Nothing will be downloaded or written beyond the generated reports (shopping-reports/).
Credentials
The skill requests no environment variables or credentials and only reads its own reference files. Required access (web fetches, occasional headless snapshots via camofox) is proportionate to its stated task of live retail research.
Persistence & Privilege
always:false (not force-included). disable-model-invocation:false is the platform default; the skill will be invocable and may spawn sub-agents during runs, which is coherent with its orchestration role. It only writes reports to shopping-reports/ and does not request system-wide changes.
如何使用
- 确保已安装 OpenClaw（本地或 Docker 部署）
- 在对话框中输入安装命令：
/install personal-shopper - 安装完成后，直接呼叫该 Skill 的名称或使用
/personal-shopper触发 - 根据 Skill 的参数说明提供必要输入，即可获得结构化输出
版本历史
v2.0.3
Security hardening: path-validated screenshot embedding, explicit domain allowlist, file access constraints, no data exfiltration rule, removed system-prompt-override patterns (renamed to task prompts), removed deterministic invocation directive, added security metadata to frontmatter
v2.0.2
Routing-logic description with USE/DON'T USE/EDGE CASES, negative examples table, worked Kill Doubt example, long-run design (compaction + continuation + artifact handoff + network containment), deterministic invocation, created brand-guideline.md, fixed html-template colors/fonts to #F8F7F4 + Rubik, fixed file paths, added store search fallback, bumped Renderer budget to 35K, cleaned dead files
v2.0.1
Fix publish path
v2.0.0
Major rewrite: English, multi-tool detection, nested sub-agents, output language choice
v1.0.0
Initial release: Diamond Search methodology with 7 agents in 3 layers for product research
元数据
常见问题
جاك العلم 是什么？
Personal Shopper — multi-agent product/service research and recommendation for Saudi Arabia. USE WHEN: - User asks to find, compare, recommend, or buy a prod... 它是一个面向 Claude Code / OpenClaw 的 AI Agent Skill 插件，目前累计下载 725 次。
如何安装 جاك العلم？
在 OpenClaw 或 Claude Code 对话框中运行命令「/install personal-shopper」即可一键安装，无需额外配置。
جاك العلم 是免费的吗？
是的，جاك العلم 完全免费（开源免费），可自由下载、安装和使用。
جاك العلم 支持哪些平台？
جاك العلم 跨平台运行，可在任意部署了 OpenClaw / Claude Code 的环境中使用（cross-platform）。
谁开发了 جاك العلم？
由 Abdullah AlRashoudi（@abdullah4ai）开发并维护，当前版本 v2.0.3。
推荐 Skills