Prompting basics
Say who, what, how and for whom. A clear prompt beats a clever one.
First exerciseTake one task you did this week and rewrite your request using the five-part checklist. Compare both answers.
Beginner
Get useful results in your first week.
Everyday work with AI
Summarise, draft, rewrite and plan. Small daily wins build the habit.
First exercisePaste a long email and ask for a three-line summary plus the one action you must take.
Checking AI output
AI can sound sure and be wrong. Learn to ask for sources and to check the important parts yourself.
First exerciseAsk a factual question, then ask the model what it is least sure about. Verify one claim against a primary source.
Privacy and safe use
Know what you should never paste into a chat, and how to use AI at work without leaking anything.
First exerciseList three kinds of data you will never paste into an AI tool, and write your own one-line rule.
Advanced
Make AI part of how you work every day.
Context and prompt chaining
Give the model the right material, then split big jobs into steps where each output feeds the next.
First exerciseTurn one long prompt into three chained steps: research, outline, draft. Note where quality improves.
Claude Skills
Package a repeatable task as a Skill so Claude does it your way every time. Skills are now an open standard (Agentic AI Foundation), so they work beyond Claude.
First exercisePick a task you repeat weekly and follow the eight-step Skills track below.
Claude Code and coding with AI
Let an AI work inside your project: read code, make changes, run tests, and explain what it did.
First exerciseInstall Claude Code, open a small project and ask it to explain the structure before changing anything.
Workflows and automation
Connect AI to the tools you already use so routine steps happen without you.
First exerciseAutomate one three-step routine, for example: new form entry, AI summary, message to yourself.
Expert
Build systems that act, retrieve and stay reliable.
Agents, tools and MCP
Give a model tools and let it decide when to use them. MCP is the open standard for connecting those tools.
First exerciseConnect one MCP server to your AI client and ask it a question that needs live data.
Retrieval and knowledge bases
Make the model answer from your documents: chunk, index, retrieve, then cite.
First exerciseBuild a tiny question-answering setup over ten of your own documents and check every answer against the source.
Evaluating and testing prompts
Write test cases before you tweak a prompt, so you know if a change really helped.
First exerciseWrite ten input and expected-output pairs for one prompt and score two versions against them.
Security, cost and reliability
Defend against prompt injection, control cost and latency by picking the right model size (for example a small, cheap model like Claude Haiku 5.5 for high-volume tasks), and plan for the day the model says something wrong.
First exerciseThreat-model one AI feature: what can untrusted text make it do? Add one guardrail and one cost limit.
Claude Skills in eight steps
A practical path from using your first Skill to maintaining a shared library.
- 1
Understand what a Skill is
A folder with a SKILL.md that tells Claude how to do one task well, plus optional scripts and files.
Read chapter one in the cookbook - 2
Use an existing Skill
Install one from the official repository and watch when Claude chooses to load it.
Browse Anthropic's Skills ↗ - 3
Read a SKILL.md closely
Notice the name, the description that triggers it, and the step-by-step instructions.
Agent Skills overview ↗ - 4
Write your first Skill
Start with a task you repeat weekly. Keep the description specific so it triggers at the right time.
Make-a-Skill chapter - 5
Add scripts and resources
Move rules, templates and small scripts into files the Skill can load only when needed.
How Skills load files ↗ - 6
Test and iterate
Run it on real tasks, note where it misses, and tighten the instructions. Repeat.
Prompt checklist for clearer instructions - 7
Share it
Put it in a repository so teammates can use the same Skill and see how it changes.
Share-a-Skill chapter - 8
Maintain a library
Review Skills on a schedule, retire the ones nobody uses and keep one owner per Skill.
Team maintenance chapter