Prompting AI Tools Effectively
Write prompts that get usable answers on the first or second try, and know why they work when they do.
This course is for members
$9 unlocks every course in the library, this one included, plus a credit to build a topic of your own and at least 10 new courses every month.
What you will learn about Prompting AI Tools Effectively
The 7-day sprint
- A Prompt Is Four Decisions, Not One Sentence. Every good prompt makes four decisions for the model: who it should be, what exact action to take, what background it needs, and what the answer should look like.
- Define Success Before You Write the Prompt. You cannot improve a prompt until you can say, in checkable terms, what a good answer would contain.
- Placement and Delimiters. Where you put material inside a prompt changes the answer, and long inputs belong near the top with your question at the end.
- Show Examples to Control Format. A couple of worked examples pins down the shape of the output far more reliably than describing the shape in words.
- Asking for Reasoning, and Knowing When It Helps. Making the model write intermediate steps before its answer helps on multi-step problems, and helps much less on everything else.
- Iterate Like an Engineer, and Set the Dials Correctly. Improve prompts by rerunning a fixed set of test cases and changing one thing at a time, and understand what the model's settings actually do.
- Context, Tools, and the Injection Problem. Once a model has files, tools, or a long conversation, the job shifts from wording a question to curating everything in the context window, and to assuming untrusted text can hijack it.
The 80/20: the 7 concepts that matter most
- Persona, Task, Context, Format
- Define Success, Then Iterate
- Put the Material First and the Question Last
- Show, Do Not Describe
- Reasoning on Demand, Not by Default
- The Model Cannot Tell Instructions From Data
- Know What the Dials Actually Do
The core idea, explained like you are 5
A model has no memory of you and no knowledge of your situation beyond the text sitting in front of it right now, so a prompt is not a command you issue, it is the context you assemble, and the answer is a continuation of that context.
Then the quiz and the ladder
10 questions that correct you when you are wrong, and beginner, intermediate, advanced levels, each with a practical exercise.
Created Sep 13, 2026. No mistakes have been reported by learners.
Written from these sources
- Prompting best practices - Claude Platform Docs, docs.anthropic.com. day 3 placement rule and the 80/20 ordering concept.
- Effective context engineering for AI agents \ Anthropic, anthropic.com. day 7 and the expert ELI5 stage.
- GitHub - anthropics/prompt-eng-interactive-tutorial: Anthropic's Interactive Prompt Engineering Tutorial · GitHub, github.com. technique ladder structure referenced in days 4 and 5.
- Anthropic’s Prompt Engineering Interactive Tutorial, simonwillison.net. day 4 typo sensitivity and day 5 scratchpad pattern.
- The Complete Prompt Engineering Guide — Anthropic | aiwithgrant, aiwithgrant.com. day 2 prerequisites and test case generation.
- GPT-4.1 Prompting Guide, cookbook.openai.com. day 7 tools field guidance.
- Prompting | OpenAI API, developers.openai.com. day 3 system versus user message placement.
- Best practices for prompt engineering with the OpenAI API | OpenAI Help Center, help.openai.com. day 6 temperature and max tokens.
- Prompt engineering best practices for ChatGPT | OpenAI Help Center, help.openai.com. day 2 iterative refinement loop.
- Prompting - Resource | OpenAI Academy, academy.openai.com. day 2 mid-conversation clarification and prompt definition.
- 5 ways to write better AI prompts for Gemini in the Workspace side panel, blog.google. day 1 four-part framework and new-teammate framing.
- Gemini for Google Workspace Prompting Guide 101, studylib.net. day 1 task verb as most important component.
- [2201.11903] Chain-of-Thought Prompting Elicits Reasoning in Large Language Models, arxiv.org. day 5 chain-of-thought origin and GSM8K result.
- [2210.03493] Automatic Chain of Thought Prompting in Large Language Models, arxiv.org. day 5 zero-shot versus few-shot CoT paradigms.
- [2506.07142] Prompting Science Report 2: The Decreasing Value of Chain of Thought in Prompting, arxiv.org. day 5 and pareto, decreasing value of chain of thought.
- [2506.14641] Revisiting Chain-of-Thought Prompting: Zero-shot Can Be Stronger than Few-shot, arxiv.org. day 4 exemplars as format alignment.
- [2406.06608] The Prompt Report: A Systematic Survey of Prompt Engineering Techniques, arxiv.org. expert stage vocabulary and taxonomy counts.
- [2302.11382] A Prompt Pattern Catalog to Enhance Prompt Engineering with ChatGPT, arxiv.org. expert stage prompt pattern catalog.
- [2311.09277] Contrastive Chain-of-Thought Prompting, arxiv.org. day 5 invalid demonstrations caveat.
- Chain of Thought Still Thinks Fast: APriCoT Helps with Thinking Slow, arxiv.org. day 5 unsettled mechanism of chain of thought.
- LLM01:2025 Prompt Injection - OWASP Gen AI Security Project, genai.owasp.org. day 7 prompt injection definition and mitigations.
- OWASP PDF v4.2.0a 20241114-202703 OWASP Top 10 for LLM Applications 2025, owasp.org. day 7 no fool-proof prevention caveat.
- OWASP Top 10 for LLM Applications (2025), aembit.io. same-channel instructions and data root cause.
- OWASP Top 10 for LLM Applications: Risks & Mitigations, mend.io. agentic consequences of successful injection.
- How AI can be hacked with prompt injection: NIST report | IBM, ibm.com. indirect prompt injection per NIST taxonomy.
- NIST AI 600-1: AI RMF Generative AI Profile Comments I. Glossary Terms:, downloads.regulations.gov. confabulation terminology in beginner stage.
- NIST AI Risk Management Framework: Agentic Profile – Lab Space, labs.cloudsecurityalliance.org. expert stage, agents judged by actions not output.
- Adaptive Evaluation of Out-of-Band Defenses Against Prompt Injection in LLM Agents, arxiv.org. expert stage, attack success rises with repeated attempts.
- Adopting NIST AI Risk Management Framework with Veeam & Securiti AI, veeam.com. logging and anomaly detection guidance in advanced exercise.