high-confidence-mingli-skill

agent
Guvenlik Denetimi
Uyari
Health Uyari
  • License — License: MIT
  • Description — Repository has a description
  • Active repo — Last push 0 days ago
  • Low visibility — Only 7 GitHub stars
Code Gecti
  • Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Gecti
  • Permissions — No dangerous permissions requested

Bu listing icin henuz AI raporu yok.

SUMMARY

高置信度命理 Skill —— 八字/紫微斗数/七政四余三盘互译的命理对话伴侣(Claude Code / Codex / Kimi 等支持 SKILL.md 的 agent CLI 通用):置信度三道题 + 斯多葛语气 + 收束自校准,支持缺盘降级,数据全本地。A Chinese metaphysics companion for agent CLIs with three-chart cross-translation (BaZi/Ziwei/Qizheng), confidence-gated interpretation, stoic voice, and self-calibrating local memory.

README.md

高置信度命理 Skill · High-Confidence Mingli Skill

八字 / 紫微斗数 / 七政四余 三盘互译的命理对话伴侣。
不是"AI 算命"——是一套带置信度体系、可自我校准的解读引擎。

A Chinese metaphysics companion for agent CLIs (Claude Code / Codex / Kimi Code — any CLI that reads SKILL.md), built on three-chart cross-translation (BaZi / Ziwei Doushu / Qizheng Siyu). Not fortune-cookie generation — an interpretation engine with a confidence system and a self-calibration loop.


为什么不一样 · Why It's Different

普通 AI 算命 本 Skill
单盘断语,张口就来 三盘互译:同一能量在八字/紫微/七政中互相验证,不叠加能量
"你本月运势不错" 置信度三道题(同领域→同方向→同名检测),⭐/⭐⭐/⭐⭐⭐ 分级标注
吉凶好坏标签 斯多葛语气:不说吉凶,说张力、火候、练习;每条判断附「你可以把握的是」
每次都像第一次见你 收束自校准:根据你的反馈持续修正解释器,越用越准
硬凑共鸣 共鸣溯源有硬性门槛:全部命题不命中就直说"不匹配"

核心机制:

  • 三套确定性排盘引擎:八字(Meeus 定气定朔,纯标准库)/ 紫微(iztro 2.6.1)/ 七政四余(自研恒星制,口径全标注)——干支、安星、躔度一律脚本输出,不靠大模型口算
  • 四模式自动路由:运势查询 / 推算推演 / 共鸣溯源 / 通用知识,开口即识别
  • 五层权重:命格 40% + 大运 20% + 时代运 15% + 流年 15% + 流月 7% + 流日 3%
  • 四层准确性防线:数据(引擎优先,口算仅作离线降级并强制标注)/ 解读(置信度三道题)/ 回归 / 活校准
  • 缺盘降级:引擎不可用时自动降级,同步封顶置信度

安装 · Install

把本仓库克隆到 Claude Code 的 skills 目录:

# macOS / Linux
git clone https://github.com/2021291696/high-confidence-mingli-skill.git ~/.claude/skills/fortune-telling

# Windows (PowerShell)
git clone https://github.com/2021291696/high-confidence-mingli-skill.git "$USERPROFILE\.claude\skills\fortune-telling"

升级:

cd ~/.claude/skills/fortune-telling && git pull

启用紫微引擎(一次性):

cd ~/.claude/skills/fortune-telling/scripts/pai_pan_ziwei && npm install

八字与七政引擎是纯 Python 标准库实现,有 python3 即可,无需安装任何包。紫微引擎不装也能用——自动降级为缺紫微盘模式(或按旧流程从排盘 App 导入)。

引擎与你的数据是分离的:git pull 只更新引擎,你的命盘和记忆存在 ~/.claude/fortune-telling/,不受升级影响,也不会被误提交。

不止 Claude Code:本 skill 是纯 Markdown 指令 + 本地文件,任何支持 SKILL.md 的 agent CLI(Codex、Kimi Code 等)都可使用——克隆到对应 CLI 的 skills 目录(如 ~/.codex/skills/fortune-telling)即可,数据目录路径不变。

使用 · Usage

在 agent CLI 中说 /fortune-telling,或直接开口(触发词自动命中):

今日运势 / 本月感情运 / 今年事业运
我和这个岗位合不合 / 该不该换城市 / 什么时候是窗口期
刷到一个说"偏印人容易过度内化"的帖子,我是不是这样?
什么是伤官配印?/ 子午冲是什么意思?

首次运行自动进入建档:引擎检测到你还没有命盘数据,会引导你完成(约 3 分钟):

  1. 出生信息(公历日期 / 时间 / 地点 / 性别)
  2. 一键跑三套确定性引擎排盘(scripts/build_chart.py):四柱大运 / 十二宫四化 / 十一曜躔度命度立命宫——不需要去文墨天机、测测等 App 抄盘
  3. 生成 chart.md(命盘)+ memory.md(记忆种子),只存在你本地

详细流程见 docs/onboarding.md。

数据与隐私 · Data & Privacy

内容 位置
引擎(可 git pull 升级) ~/.claude/skills/fortune-telling/
你的命盘 chart.md ~/.claude/fortune-telling/
你的记忆 memory.md + 归档 ~/.claude/fortune-telling/

所有个人数据只存在本地,引擎本身不联网、不上传、不绑定任何账户。

目录结构 · Structure

├── SKILL.md                  # 引擎主文件:路由/置信度/语气/权重/降级推算规则
├── scripts/                  # 三套确定性排盘引擎(v4 新增)
│   ├── pai_pan_bazi.py       # 八字:四柱/大运/神煞(Meeus 定气,纯标准库)
│   ├── pai_pan_qizheng.py    # 七政:躔度/命度/立命宫/恩用难仇(自研恒星制,口径声明在文件头)
│   ├── pai_pan_ziwei/        # 紫微:十二宫/四化/大限/格局(iztro 2.6.1,需 npm install)
│   ├── build_chart.py        # 建档总控:出生信息 → 三引擎 → chart.md
│   └── test_*.py             # 各引擎回归测试(对拍黄金值 + 古典规则例)
├── workflows/                # 五个执行流程(路由命中后读取执行)
│   ├── fortune-daily-weekly-monthly.md   # 运势查询
│   ├── fortune-specific.md               # 推算推演
│   ├── fortune-resonance.md              # 共鸣溯源
│   ├── fortune-knowledge.md              # 通用知识
│   └── fortune-close.md                  # 收束校准
├── templates/                # chart.md / memory.md 模板
├── docs/onboarding.md        # 首次建档流程
└── LICENSE                   # MIT

限制 · Limitations

  1. 七政四余月限/小限需限度推算,当前不包含
  2. 七政盘制为恒星制(角宿一=0°),与回归制排盘软件相差约 24°,属学派差异——口径与出处见 scripts/pai_pan_qizheng.py 文件头
  3. 紫微引擎需要 node(npm install 一次);无 node 时该盘自动降级,其余两盘不受影响
  4. 三元九运交接点存在学派争议(本引擎统一用 2004/2024)
  5. 运势是概率框架,不是预言

English Overview

High-Confidence Mingli Skill is a Claude Code skill that turns your agent into a Chinese metaphysics reading companion. Instead of single-system fortune-telling, it cross-translates three chart systems — BaZi (Four Pillars), Ziwei Doushu (Purple Star Astrology), and Qizheng Siyu (Seven Governors & Four Remainders) — treating them as three languages describing the same underlying energy, never stacking them.

Key mechanics:

  • Confidence-gated interpretation: every cross-chart correlation must pass three checks (same domain → same direction → same source, with a same-name trap detector) before it earns ⭐/⭐⭐/⭐⭐⭐ confidence markers
  • Four auto-routed modes: time-based fortune queries, specific-question divination, resonance tracing ("I saw a post that says..."), and general knowledge Q&A
  • Stoic voice: no "auspicious/inauspicious" labels — tensions, timing, and one concrete practice suggestion per reading
  • Self-calibrating memory: each session closes by comparing its judgments against your feedback and refining the interpreter profile stored locally
  • Graceful degradation: works with BaZi alone; missing charts cap the confidence ceiling accordingly
  • Privacy by design: engine and user data are fully separated — your chart lives outside the skill directory and is never uploaded

Install: clone this repo into ~/.claude/skills/fortune-telling, then say /fortune-telling in Claude Code. First run walks you through a ~3-minute onboarding (birth data → all three charts computed by the built-in deterministic engines — BaZi via Meeus-based solar-term math in pure stdlib Python, Ziwei via iztro 2.6.1, Qizheng via a self-built sidereal engine with fully documented school conventions; no app copy-pasting needed). All data stays local under ~/.claude/fortune-telling/.

Note: interpretation content is primarily in Chinese, as the source metaphysics tradition is Chinese.

License

MIT — see LICENSE.

Yorumlar (0)

Sonuc bulunamadi