Prompt Lab
Write real prompts, system prompts, tool definitions, CLAUDE.md files and skills. Claude tries your work on test messages, then grades it against a rubric and tells you what to change.
- 1Write your prompt, tool or document from the task.
- 2Claude tries it on the test messages.
- 3A grader scores it point by point and gives you one next step.
Meet Claude
- PromptTurn a vague request into a good prompt
Practises: Getting better answers
Open exercise - DocumentWrite instructions for a project
Practises: Projects, artifacts and skills
Open exercise
Working well with AI
- DocumentDecide what to hand to AI
Practises: Delegation: what to hand to AI
Open exercise - PromptDescribe the product, the process and the behaviour
Practises: Describing well and judging the answer
Open exercise - DocumentJudge an AI answer
Practises: Describing well and judging the answer
Open exercise - DocumentOwn what AI helped you make
Practises: Diligence: owning what AI helps you make
Open exercise
AI fluency for your world
- DocumentAn assignment where AI use is visible
Practises: AI fluency for teachers and trainers
Open exercise - DocumentData rules for a nonprofit team
Practises: AI fluency for nonprofits
Open exercise - System promptA reply assistant for a kirana store
Practises: AI fluency for small businesses
Open exercise - System promptA tutor that helps you think
Practises: AI fluency for students
Open exercise
Claude at work
- DocumentHand Cowork a whole task
Practises: Handing off tasks with Cowork
Open exercise - DocumentInstructions for Claude in a team channel
Practises: Shaping Claude in your channels
Open exercise - DocumentA rollout decision memo
Practises: Rolling out Claude to a company
Open exercise
Claude Code
- DocumentWrite a CLAUDE.md
Practises: CLAUDE.md: teaching Claude your project
Open exercise - DocumentPlan your hooks
Practises: Permission modes and hooks
Open exercise - DocumentDefine a subagent
Practises: Subagents
Open exercise - DocumentWrite a skill
Practises: Agent skills
Open exercise - DocumentA safe unattended run
Practises: Headless runs, CI and unsupervised work
Open exercise
Build with the Claude API
- System promptA system prompt for a help bot
Practises: Conversations, system prompts and streaming
Open exercise - System promptGet clean data out
Practises: Getting structured data back
Open exercise - System promptClear, structured and with examples
Practises: Prompt engineering techniques
Open exercise - DocumentDesign an eval
Practises: Testing prompts with evals
Open exercise - Tool definitionDefine a tool
Practises: Tool use: letting Claude call your code
Open exercise - Tool definitionA tool with useful filters
Practises: Many tools, many turns
Open exercise - System promptAnswer only from the documents
Practises: Retrieval: answering from your documents
Open exercise - DocumentWhen to use extended thinking
Practises: Extended thinking
Open exercise - DocumentLay out a prompt for caching
Practises: Prompt caching
Open exercise - System promptA routing prompt
Practises: Workflows: chaining, routing and parallel work
Open exercise - DocumentRules for an agent
Practises: Agents: tools, feedback and when to use them
Open exercise - DocumentMake computer use safe
Practises: Computer use
Open exercise
Model Context Protocol
- Tool definitionA tool for your MCP server
Practises: Building an MCP server
Open exercise - DocumentTool, resource or prompt
Practises: Resources and prompts
Open exercise