
Compose a prompt library for your domain
You'll have a working document with 8–12 tested prompts that reliably produce accurate, low-hallucination outputs for the specific tasks you do most often — medicine, law, finance, coding, or another
What you'll make
Domain Prompt Library
- High-Stakes Prompts: 4-6 prompts for mission-critical tasks with validation data
- Routine Prompts: 4-6 prompts for daily tasks with reliability scores
- Testing Protocol: your method for validating new prompts and models
- Failure Cases: documented instances where prompts failed and why
Submit your prompt library as a structured document with actual prompts, test results, and your validation methodology
The 4 parts
Each part ends with a check you can see.
Audit Your High-Stakes Tasks
Identify the 10-15 AI tasks you do where wrong answers have real consequences
Done when
A ranked list of your most frequent AI tasks scored by consequence-of-error (1-5 scale) with at least 3 tasks scoring 4-5
Run Your Own Model Comparison
Test the same 5-8 prompts across different models using facts you can verify in your domain
Done when
Comparative data showing hallucination rates for 5+ domain prompts tested on at least 2 different AI models, with documented fact-checking methodology
Optimize Your Best-Performing Prompts
Take your most reliable prompts from testing and refine them to consistently produce accurate outputs
Done when
8-12 refined prompts with documented accuracy improvements and specific modification notes explaining what changes reduced hallucinations
Document Your Validated Library
Compile your tested and refined prompts into a reusable library with usage notes and reliability data
Done when
Complete prompt library document with 8-12 prompts, each including the prompt text, use case, accuracy data, and modification notes
Capstone: you put together your Domain Prompt Library.
Where this helps
An elective: a single lesson that sits outside rtwo's job certificates and doesn't count toward one.
About this lesson
What you'll be able to do
- Run 8-12 tested prompts that produce reliable outputs in your domain with measurably lower hallucination rates
- Identify the 3-5 prompt patterns where newer models consistently outperform older ones in your specific work
- Have a documented testing protocol you can run whenever new AI models are released
Who it's for
- Professionals in high-stakes domains (medicine, law, finance, coding) who need reliable AI outputs
- Anyone who's been burned by AI hallucinations and wants systematic ways to reduce them
- Knowledge workers who use AI regularly but haven't formalized their prompt testing
What it won't cover
- General prompt engineering theory or techniques
- Model fine-tuning or custom training
- Prompts for creative writing or brainstorming
- Technical implementation of hallucination detection systems
Skills you'll practise
- Domain-specific hallucination testing
- Prompt validation against known facts
- Model comparison on professional tasks
- Systematic prompt refinement