Yangjin Xuan Codex Skill

yangjin-xuan is a public Codex Skill for Xuanxue culture interpretation and content production workflows.

It is designed for:

• Yijing / Zi Wei Dou Shu /命理文化解读
• Local corpus-assisted research
• Metaphysics-themed Xiaohongshu posts
• Short-video scripts
• Livestream talking points
• Knowledge-base organization

It is not a fortune-telling promise, not a deterministic prediction system, and not a substitute for medical, legal, investment, or other professional advice.

What Is Included

yangjin-xuan-skill/
├── SKILL.md
├── README.md
├── references/
│   ├── method.md
│   ├── content_style.md
│   ├── safety_boundary.md
│   ├── openclaw_usage.md
│   └── public_sources.md
├── openclaw/
│   ├── agent_prompt.md
│   ├── quickstart.md
│   └── test_prompts.md
├── scripts/
│   └── query_local_corpus.py
├── examples/
│   ├── input_example.md
│   └── output_example.md
└── .gitignore

What Is Not Included

This public Skill intentionally excludes:

• Copyrighted books, PDFs, EPUBs, scans, and full OCR text
• Private local corpora
• Generated web indexes from private books
• API keys and .env files
• User records, reading history, private notes, or databases
• Local absolute paths from the original working machine

Install In Codex

Copy this folder into your Codex skills directory:

mkdir -p ~/.codex/skills
cp -R yangjin-xuan-skill ~/.codex/skills/yangjin-xuan

Then start a new Codex session and ask for a task such as:

用 yangjin-xuan 帮我把“紫微斗数四化”做成一个小红书选题和短视频脚本。

Try In OpenClaw Or Another Agent Platform

If your OpenClaw setup supports installing a Skill from a GitHub URL, use:

https://github.com/yjin5562-dev/yangjin-xuan-skill

If it requires a local Skill folder, clone the repository first:

git clone https://github.com/yjin5562-dev/yangjin-xuan-skill.git

Then import or copy the cloned folder according to your OpenClaw Skill/plugin workflow.

For platforms that ask for a system prompt or agent instruction, paste the content of:

openclaw/agent_prompt.md

For quick testing, use:

openclaw/test_prompts.md

Suggested test prompts:

用 yangjin-xuan 做一次易经风格的文化解读：我明天要见一个合作方，想知道沟通上要注意什么。请不要做确定性预测，要给白话解释、风险清单和行动建议。

用 yangjin-xuan 帮我生成一个“紫微斗数不是算准，而是建立文化解读工作流”的小红书脚本。

This Skill does not include a built-in fortune-telling engine, private book corpus, bazi calculator, Zi Wei charting engine, or API key. It provides the workflow, safety boundary, content structure, and optional local-corpus search entry point.

Without any private corpus or extra charting library, it can still run as a ready-to-use cultural consultation and content-production agent powered by the model configured in OpenClaw.

Local Use

If you have a private local corpus, keep it outside the public Skill repository. Then run:

python3 scripts/query_local_corpus.py "四化 化忌 夫妻宫" --root ~/xuan-private-corpus

Or set an environment variable:

export XUAN_CORPUS_ROOT=~/xuan-private-corpus
python3 scripts/query_local_corpus.py "命宫 身宫 紫微"

The script returns short snippets and file paths. It does not upload anything.

Connect To A Local Model Or App

This Skill can support a local app or model by providing:

• source-search snippets from private files,
• safety boundary instructions,
• content-writing structure,
• interpretation templates,
• source discipline for citations.

Recommended pattern:

1. Keep private books and OCR text in a separate local-only folder.
2. Use scripts/query_local_corpus.py to retrieve small snippets.
3. Pass only necessary snippets to the model.
4. Keep generated public content citation-light unless the source is verified public-domain or user-owned.

GitHub Release Checklist

Before publishing:

• No API keys or .env files.
• No private personal information.
• No copyrighted book full text, PDFs, scans, EPUBs, or OCR dumps.
• No generated indexes derived from copyrighted private books.
• No private user records or app databases.
• No original-machine absolute paths in public docs, unless clearly marked as examples to replace.
• SKILL.md has name and description frontmatter.
• scripts/query_local_corpus.py works with a separate local corpus path.

License

Choose a license before publishing. If unsure, use MIT for the Skill code and docs, while keeping private corpora excluded.