Prompting Claude: Instructions, Context and Testing

What's inside

1. Level 1: Course Ground Rules and How Claude Reads TextFree
   Learn that this course is independent of Anthropic, how to verify fast-changing Claude features on Anthropic's current pages, and how Claude processes a prompt as tokens.
2. Level 2: Context Window and Non-DeterminismFree
   Learn what the context window contains, why it is finite, and why the same prompt can produce different answers.
3. Level 3: Clear, Direct Instructions with Reasons
   Learn to write specific, unambiguous instructions and to state the reason behind each requirement.
4. Level 4: Role and Audience
   Learn to set the role Claude should take and the audience it is writing for so that tone, depth and vocabulary fit the task.
5. Level 5: Zero-, One- and Few-Shot Examples
   Learn when to rely on instructions alone and when to add one or several examples to show the pattern required.
6. Level 6: Structuring Prompts with Tags and Sections
   Learn to separate instructions, context, examples and input using tags or labelled sections so that each part is read correctly.
7. Level 7: System Versus User Placement
   Learn which content belongs in standing instructions and which belongs in the individual message, first in chat and then in the API.
8. Level 8: Output Constraints and Formats
   Learn to control the length, format, tone and layout of a response through explicit output specifications.
9. Level 9: Step-by-Step Reasoning in the Prompt
   Learn to ask Claude to reason through a problem in visible steps before answering, and to keep the reasoning separate from the final answer.
10. Level 10: Extended and Adaptive Thinking, and Effort
    Learn how Claude's built-in thinking modes and effort levels differ from prompted reasoning, and how to set them in chat and in the API.
11. Level 11: Structured Output, Schemas and Validation
    Learn to obtain machine-readable output that conforms to a schema and to validate it before use.
12. Level 12: Decomposition and Prompt Chaining
    Learn to split a complex task into a sequence of focused prompts in which each output feeds the next.
13. Level 13: Long Documents: Placement and Quoting
    Learn where to place long documents in a prompt, how to label them, and how to have Claude extract relevant quotes before answering.
14. Level 14: Grounding Answers and Reducing Hallucination
    Learn to tie answers to supplied sources and to reduce unsupported claims through explicit permission to say the answer is unknown.
15. Level 15: Context Engineering: Curate, Summarise, Compact
    Learn to decide what to include in the context, what to leave out, and when to summarise or compact a long conversation to counter drift and bloat.
16. Level 16: Memory and Persistent Instructions
    Learn how instructions and facts can persist across conversations, and how to write and maintain them so they stay accurate.
17. Level 17: Reuse: Templates, Projects, Skills and Caching
    Learn to reuse proven prompts and context through templates with variables, Projects, Skills and prompt caching.
18. Level 18: Prompt Injection and Untrusted Input
    Learn how instructions hidden in documents, web pages or user input can redirect Claude, and how to reduce that risk.
19. Level 19: Testing Prompts and Diagnosing Failures
    Learn to build a small test set, compare prompt variants, improve from failures, and re-test when the model or release changes.
20. Level 20: Final Challenge: Rebuild and Test a Flawed Prompt Exam
    Apply every technique from the course to diagnose, rebuild and test a flawed prompt and the context supplied with it.

Access

The first 2 levels are free with a free account. Every level, the podcast edition and the AI tutor come with All Access at £4.99/month or any Creator plan — see pricing.

Start free →