AI Fortune Telling That Doesn't Hallucinate: This Skill Uses Real Algorithms Behind the Scenes
👤 Who This Is For
The biggest problem with AI fortune telling isn't that it's "superstitious" — it's that it's inaccurate. Pure LLMs frequently mess up the Four Pillars calculation, misidentify the Day Master, or get the pattern wrong. One wrong step, and the entire reading falls apart. This Skill takes a different approach: let algorithms handle the calculation, and let the AI handle the interpretation.
Three Things It Does
**First, algorithm-powered charting.** The Eight Characters (Bazi) pillars, Ziwei Doushu's twelve palaces, and major/minor fortune periods are all calculated by a built-in algorithm library — specifically, the open-source Yiqi project (MIT licensed), empirically verified for accuracy. The LLM only interprets; it doesn't calculate.
**Second, pattern augmentation.** On top of the raw chart, the Skill adds a layer of traditional pattern analysis —格局 (pattern), 旺衰 (strength), 调候 (temperature adjustment), and various compatibility/incompatibility rules — then feeds this structured context to the LLM for grounded analysis.
**Third, cross-verification.** This is the project's unique selling point: it runs both Bazi and Ziwei Doushu independently, then cross-checks their conclusions. Do the two systems point to the same life themes? Are the timing windows aligned? When they conflict, which one wins? You get a nuanced reading instead of a single-system guess.
What Using It Feels Like
Once installed, just tell your Agent your birth details: "I was born on January 1, 2000 at 12:00 PM. Can you read my chart?"
Three analysis modes: Bazi-only, Ziwei-only, or combined cross-verification. Two output formats: a deep-dive Markdown report, and a single-file HTML poster with ink-wash aesthetics — perfect for sharing on social media.
Everything runs locally. No internet required for charting. Output files are gitignored by default. Privacy was clearly a design consideration.
Why It's Trending
Chinese metaphysics + AI is a consistently viral combination, but this project stands out because it actually works. 500 stars in 12 days is notable for a niche SKILL.md project. The ink-wash poster feature probably helped — it's visually distinctive and highly shareable.
If you're building an Agent that serves Chinese-speaking users, this is a drop-in Skill that adds a genuinely useful (and highly differentiated) capability. The SKILL.md standard means it works across Claude Code, Codex, Cursor, and OpenClaw out of the box.
Get Started
Open Source · Commercial Friendly
git clone https://github.com/dzcmemory-web/bazi-ziwei-skill.git && cd bazi-ziwei-skill/calculator && npm install