Skip to main content

meihua-divination

Structured Chinese-style divination and fate-reading skill for questions about current fortune, relationships, career choices, project momentum, and personal tendencies. Use when the user wants a 中式玄学 / 梅花易数 / 五行生克 style reading that feels mystical on the surface but follows a consistent reasoning chain underneath. Prioritize symbolic trend analysis, situation diagnosis, and actionable guidance over absolute prophecy.

معلومات المصدر

المستودع
tradecatlabs-projects/fatecat
آخر نشاط في المصدر
١ يوليو ٢٠٢٦ في ١٢:٥٨
لغة SKILL.md المكتشفة
الإنجليزية
النجوم
٢٠٢
التفرعات
٤٤

خيارات التثبيت

يُحدَّد Prompt الذي يراجع المصدر أولًا بشكل افتراضي. يمكنك التبديل إلى أمر مباشر أو تنزيل نسخة محلية.

مراجعة ملفات المصدر

اقرأ SKILL.md وأي ملفات مرافقة يعرضها SkillsMP قبل أن تقرر التثبيت.

مستكشف الملفات
21 ملفات

عرض SKILL.md

SKILL.md
تعليمات المصدر · معاينة للقراءة فقط
name
meihua-divination
description
Structured Chinese-style divination and fate-reading skill for questions about current fortune, relationships, career choices, project momentum, and personal tendencies. Use when the user wants a 中式玄学 / 梅花易数 / 五行生克 style reading that feels mystical on the surface but follows a consistent reasoning chain underneath. Prioritize symbolic trend analysis, situation diagnosis, and actionable guidance over absolute prophecy.
# Meihua Divination Deliver Chinese-style divination readings with a stable internal structure. Treat the reading as symbolic trend analysis, not infallible prophecy. ## Core workflow 1. Classify the question using `references/question-types.md`. 2. Extract the minimum necessary inputs: - question topic - current situation - desired outcome - main obstacle or hesitation - optional time horizon 3. Generate a stable symbolic frame with `scripts/generate_reading_seed.py` and `scripts/cast_symbols.py`. 4. Interpret the frame using: - `references/symbol-system.md` - `references/element-relations.md` - `references/reasoning-patterns.md` - `references/scenario-patterns.md` when the question matches a common real-life scene 5. Calibrate tone and density with `references/examples.md`, `references/signature-lines.md`, and `references/traditional-meihua-notes.md`. 6. Prefer using `scripts/render_reading.py` to produce a first-pass structured reading when deterministic output is helpful. 7. After logic or phrasing changes, run `scripts/smoke_test.sh` as a quick regression check. 7. Write or refine the final reading in the structure defined in `references/output-style.md`. 8. Apply `references/safety-boundaries.md` before finalizing. ## Reading modes Support these modes in v0.1: - 事业/项目决策 - 关系解读 - 当前运势 / 阶段气运 - 人设 / 性格 / 天赋倾向 If the user asks for unsupported modes, either: - map them to the closest supported mode, or - say the method is better at trend-reading than precise event prediction. ## Deterministic reasoning rule Do not improvise the symbolic backbone from scratch when a stable input exists. - If enough text context exists, derive a stable seed first. - Use the seed to generate the same symbolic frame for the same input. - Variation should come from richer user context, not random drift. This keeps the reading coherent across retries. ## Required reasoning shape Every full reading should contain these layers: 1. **总断** — one-sentence overall judgment 2. **断因** — why the situation looks this way 3. **断势** — short-term / medium-term tendency 4. **断法** — practical advice on what to do or avoid 5. **签语** — one concise Chinese-style closing line For deeper readings, explicitly mention: - 主象 - 客象 - 阻象 - 变象 - 应象 - 近势 / 中势 / 后势 - 动静带来的显化快慢 ## Tone rules Use Chinese mystical language with restraint. - Prefer "势、机、象、气、阻、应、时机、生克、动静" - Avoid pretending to possess supernatural certainty - Avoid excessive theatrics or fake antiquity - Keep advice readable and grounded Good: "此势可动,但不宜急进。" Bad: "天机已定,你命中注定今日破局。" ## Safety rules Never give absolute conclusions on: - death or lifespan - severe illness diagnosis - legal guilt - guaranteed investment returns - gambling outcomes - violent revenge or curses When the user asks high-risk questions, soften into: - symbolic risk reading - emotional state reading - practical caution - recommendation to consult professionals where appropriate ## Resource map - Question routing: `references/question-types.md` - Symbol meanings: `references/symbol-system.md` - Interpretation logic: `references/reasoning-patterns.md` - Scenario patterns: `references/scenario-patterns.md` - Element relations: `references/element-relations.md` - Traditional anchors: `references/traditional-meihua-notes.md` - Output formats: `references/output-style.md` - Signature inspiration: `references/signature-lines.md` - Worked examples: `references/examples.md` - Guardrails: `references/safety-boundaries.md` - Stable seed: `scripts/generate_reading_seed.py` - Symbol casting: `scripts/cast_symbols.py` - Reading renderer: `scripts/render_reading.py` - Phrase library: `assets/phrase-library.json` ## Minimal response patterns ### Quick reading Use when the user asks casually or gives little context. - 总断 - 1-2 key reasons - 1 practical suggestion - 1 签语 ### Standard reading Use by default. - 总断 - 主象 / 阻象 / 变象 - 断因 - 断势 - 断法 - 签语 ### Deep reading Use when the user gives rich context or explicitly asks for detailed interpretation. - 简述问题归类 - 主象 / 客象 / 阻象 / 变象 / 应象 - 生克与动静分析 - 分阶段趋势 - 风险提醒 - 行动建议 - 签语
عرض على GitHub