What it does

yao-open-skills is a curated, bilingual repository of production-ready AI skills—structured prompt-and-workflow packages for tasks like security auditing, business analysis, reading-data visualization, and tutorial generation. Each skill ships with intake logic, evaluation criteria, multi-format report pipelines, and explicit boundaries on what it won’t do.

The interesting bit

The project treats prompts as assets, not tricks. Every skill must pass a “publication standard”: clear topic, reusable without private context, sanitizable of sensitive data, and maintained by a human. There’s even a meta-skill (yao-meta-skill) for creating skills—a rare case of prompt-engineering eating its own tail.

Key highlights

• Yao Crux Skill: Diagnostic reports on complex problems using contradiction analysis, Bayesian evidence updating, and resource-allocation recommendations; outputs Markdown, HTML, DOCX, PDF, and JSON.
• Yao WeRead Skill: Generates 26-chart HTML reports from WeChat Reading data, with API keys read only from environment variables and real user reports kept out of the repo.
• Yao Websecurity Skill: 275-check vulnerability ontology across 11 risk domains, with five review modes from static to dynamic-active; scrubs paths, tokens, and keys before rendering.
• Yao Tutorial Skill: End-to-end tutorial production from topic or source packet through research, outline, and multi-format output with illustrations.
• Registry, naming conventions, and sync scripts keep the collection maintainable as it grows.

Caveats

• Featured “Skill Doctor / Optimizer / Ranker” lines are described as product directions, not yet implemented skills.
• The README is truncated mid-sentence in the Yao Tutorial Skill section, so full capabilities there are unclear.
• Most skills are documented in Chinese with English READMEs; the bilingual split may slow navigation.

Verdict

Worth browsing if you build prompt libraries for teams and want governance patterns—registry design, sanitization rules, versioned publication—rather than one-off prompt hacks. Skip if you’re looking for drop-in English-language tools or immediate API integrations; these are structured methodologies, not packaged software.