← Back to Skills Marketplace
575
Downloads
0
Stars
2
Active Installs
2
Versions
Install in OpenClaw
/install geekbench
Description
Provides real-time searches and verified Geekbench scores for the latest officially released flagship smartphones by brand.
README (SKILL.md)
Geekbench 技能
核心原则
- 不要依赖模型知识：模型训练数据有时效性，最新产品信息需实时搜索
- 多方信源验证：通过搜索获取最新发布信息
- 实时确认：每次查询都重新搜索确认
确定最新旗舰流程
Step 1: 搜索最新信息
- 搜索词: "品牌 + 最新旗舰 + 发布" 或 "品牌 + 2025 旗舰"
- 示例: "OPPO 最新旗舰 发布"
- 目标: 获取最新已发布的产品名称
Step 2: 验证发布时间
- 确认产品是否已正式发布
- 排除"即将发布"、"谍照"、"传闻"等未发布信息
Step 3: 搜索跑分
- 使用 "产品名 + geekbench" 搜索
- 获取内部型号
- 用内部型号查询详细跑分
Step 4: 分析验证
- 按Geekbench版本分组统计
- 验证数据与发布时间的一致性
- 标记异常数据（测试版本、样本过少等）
示例：OPPO Find X9 系列
错误做法（依赖模型知识）
- 模型知识: Find X8 (2024年发布)
- 结果: 返回过时信息
正确做法（实时搜索）
- 搜索 "OPPO 最新旗舰 发布"
- 找到: Find X9 系列（2025年发布）
- 搜索 "OPPO Find X9 geekbench"
- 获取内部型号和跑分数据
常见内部型号格式
- OPPO: 需搜索获取Find X9型号
- 荣耀: 需搜索获取Magic 9型号
- 小米: 需搜索获取小米18型号
- 苹果: 需搜索获取iPhone 17/18型号
注意事项
- 不同Geekbench版本分数不可直接比较
- 优先使用 Geekbench 6.5.0 数据
- 检测 Single-Core > 3000 为高分设备
- 未发布产品的跑分可能是早期测试数据
Usage Guidance
This skill appears to deliver what it promises (geekbench scraping and monitoring), but there are red flags you should address before installing: (1) Review the three bundled Python files (geekbench_crawler.py, monitor.py, tasks.py) to confirm they only access expected endpoints (e.g., browser.geekbench.com and search engines) and do not POST data to unknown servers or read local secrets. (2) Ask the publisher for the source repository or homepage and for a clear runtime/install/permission list (Python version, required packages). (3) If you must run it, execute it in a sandboxed environment and inspect outgoing network connections. (4) Prefer skills that explicitly declare required binaries/dependencies and a verifiable source; if you cannot audit the code, treat this skill as higher risk.
Capability Analysis
Type: OpenClaw Skill
Name: geekbench
Version: 1.0.1
The skill bundle is designed for scraping and analyzing Geekbench scores, saving reports locally, and sending notifications via the `openclaw message send` command. All network requests are directed to `https://browser.geekbench.com`, and file system operations are confined to the designated OpenClaw workspace (`/Users/ding/.openclaw/workspace/geekbench`). The `SKILL.md` instructions are clear and functional, guiding the AI agent on how to perform its task without any signs of prompt injection attempts. There is no evidence of data exfiltration, malicious execution, persistence, or obfuscation.
Capability Assessment
Purpose & Capability
The name/description (fetch verified Geekbench scores) aligns with the presence of a crawler, monitor, and archived data. However, the skill metadata claims 'instruction-only' / no install spec and lists no required binaries while shipping three substantial Python scripts (geekbench_crawler.py, monitor.py, tasks.py). That is inconsistent: bundled code implies the need for a runtime (Python) and network access but those requirements are not declared.
Instruction Scope
SKILL.md correctly instructs the agent to perform live web searches and verify Geekbench results; it does not request reading local secrets or unrelated files. The runtime instructions are narrowly scoped to search/verify tasks. However, the shipped monitor/crawler code and archived reports indicate automated scraping/monitoring behavior not documented in SKILL.md — the agent’s instructions and the included automation artifacts are not fully reconciled.
Install Mechanism
No install spec is provided (lower install risk), but code files are included and marked for execution. There is no declaration of how to run them (no required binaries listed, no Python requirement). This mismatch increases risk because bundled scripts could be executed in the agent environment without the user knowing required runtime, dependencies, or network behavior.
Credentials
The skill requests no environment variables or credentials (proportionate for a public data scraper). That is good. Still, because the code is present but not documented, there's no guarantee the code won't attempt to read environment variables or local files at runtime — the metadata doesn't declare any such accesses, so a code review is needed to confirm proportionality.
Persistence & Privilege
always:false and model invocation not disabled (defaults). The skill does not request permanent/system-level presence. There is no evidence it modifies other skills or agent-wide settings in the provided artifacts.
How to Use
- Make sure OpenClaw is installed (local or Docker)
- Run the install command in chat:
/install geekbench - After installation, invoke the skill by name or use
/geekbench - Provide required inputs per the skill's parameter spec and get structured output
Version History
v1.0.1
No changes detected in this version.
- No file changes were made from the previous version.
v1.0.0
- Initial release of the Geekbench skill.
- Implements a real-time, multi-source workflow for confirming the latest flagship models and their Geekbench scores.
- Clearly distinguishes between released and unreleased products, excluding rumors and leaks.
- Guides users to verify product launch status and search for internal model numbers for accurate benchmarking.
- Emphasizes using up-to-date Geekbench data, prioritizing Geekbench 6.5.0, and highlights methods to identify anomalous or early test scores.
Metadata
Frequently Asked Questions
What is Geekbench?
Provides real-time searches and verified Geekbench scores for the latest officially released flagship smartphones by brand. It is an AI Agent Skill for Claude Code / OpenClaw, with 575 downloads so far.
How do I install Geekbench?
Run "/install geekbench" in the OpenClaw or Claude Code chat to install it in one step — no extra setup required.
Is Geekbench free?
Yes, Geekbench is completely free (open-source). You can download, install and use it at no cost.
Which platforms does Geekbench support?
Geekbench is cross-platform and runs anywhere OpenClaw / Claude Code is available (cross-platform).
Who created Geekbench?
It is built and maintained by d404 (@dingqing404); the current version is v1.0.1.
More Skills