Yao GEO Skills

Install

Works in

• Claude Code
• Codex
• Koda
• Coddy

Manual install

git clone https://github.com/yaojingang/yao-geo-skills.git

What it does

Yao GEO Skills is a repository of 24 skills for GEO, meaning optimization of a site and brand for neural-network answers. The skills are grouped into seven areas: strategy, technical page review, content production, knowledge base and brand graph, measurement, query research and work with the GEOFlow system. Each skill sits in its own folder under skills/ with a SKILL.md file, a manifest, a brief template and trigger checks. The `yao-geo-page-audit` skill takes a site address, picks the homepage and one first-level and one second-level page, and checks the response code, robots, sitemap, canonical, headings, Schema markup and whether the main text arrives in the initial HTML. The output is a list of code and text fixes, each with a priority, an owner and an acceptance test. The content skills write titles, explainers, comparisons and rankings, and `yao-geo-article-friendly` rewrites a finished article for citation. Most reports come in four formats: Word, PDF, HTML and Markdown. The measurement skills ask the web versions of ChatGPT, DeepSeek and Doubao the same questions repeatedly through `OpenCLI Browser Bridge` (Doubao also has an Android emulator and Appium route), then count how often the target brand and its competitors are mentioned, their position in the answer and the sources. The research skills collect Baidu, Xiaohongshu and Douyin search suggestions two levels deep, save screenshots and export an Excel table.

Who it is for. For SEO specialists, marketers and agencies that prepare a site and content for neural-network answers and hand clients Word and PDF reports, and for teams running a GEOFlow system.

Example request

Run a GEO audit of our site's homepage and give me a fix list for the developer and the editor with priorities, saved as Word and PDF

Limitations

• The skills are tuned to the Chinese ecosystem. The panorama audit looks at DeepSeek, Doubao, Qwen, Kimi and Yuanbao, suggestions are collected from Baidu, Xiaohongshu and Douyin, and the reports and examples in the repository are in Chinese.
• Sampling needs Node.js 18 or newer, Python 3.10 or newer, OpenCLI 1.8.4 or newer with the Browser Bridge extension and a Chrome or Edge profile that is already signed in. The skill does not sign in, solve captchas or bypass checks. The default pause between queries is random, from 30 seconds to a minute.
• Sampling figures are estimates from repeated samples, not exact shares. The skill infers competitors with a heuristic, so check the aliases before sending a report to a client.
• During suggestion collection a captcha or a login prompt stops the batch and the term is marked as blocked. The platform may limit even typing into its search box. By default the skill runs up to 200 queries with 10 suggestions each.
• `yao-geoflow-template` stays as an entry point for old GEOFlow PHP templates, while Laravel Blade themes go through `yao-geoflow-design`. License MIT.

How to disable. Remove the skill folders from your agent's skills directory and the cloned repository. If you set up sampling, disable the Browser Bridge extension in Chrome or Edge and sign out of the AI accounts in that profile. Reports, JSON exports and screenshots stay in the output folders you chose, so delete them by hand.

README in short

The repository's front page is written in Chinese, and an English version sits in docs/README.en.md. It has a table of seven groups with the number of skills in each, a description of every skill, three ways to get the code (clone, ZIP, a partial checkout of one skill) and the repository principles: one skill solves one task, has clear boundaries, and examples are stripped of private data. Many skills ship public examples in Markdown, HTML, Word and PDF. The skill list lives in registry/skills.json, the structure is checked with python3 scripts/validate_repository.py, and a navigation page sits in index.html.