Most people prompt AI the way they Google — short, vague, no structure. Then wonder why the output is useless.

This library fixes that. 50+ production-tested Few-Shot and Chain-of-Thought templates across three domains where bad AI output has real consequences: legal drafting, software development, and academic research.

What's inside:

Legal (L-01 to L-05) — NDA clauses, contract risk analysis, IRAC research summaries, termination letters, Terms of Service generation. Every template includes anti-hallucination guardrails and citation discipline rules built in.

Code (C-01 to C-06) — Function implementation with test cases, security-focused code review, API integration boilerplate, SQL optimization, system design, regex builder. Specify your language and version, get production-grade output.

Research (R-01 to R-05) — Literature reviews, research question refinement using PICO framing, statistical results interpretation, IMRAD abstracts, peer review response letters.

Advanced techniques — Self-critique loops, Constitutional checking, meta-prompting, prompt chaining architecture.

Works on ChatGPT (GPT-4o) and Claude. Includes a full guide on tuning each template per model.

Tested on GPT-4o and Claude Opus/Sonnet. Copy, paste, customize.