Phase 3 · Claude · Level 2 · Practitioner
Claude: Practitioner Quiz · 10 min
Why it matters A quick check of your judgement before you move on. There's no penalty for a low score, and you can retake it, or take it first to test out of a level you already know.
These scenarios test the practitioner judgement calls: when to build a Project, how to work an Artifact, making Claude actually calculate, controlling what it remembers, and verifying a cited report. They're not trivia; they're the decisions you'll face doing real work in Claude. Pick the best option for each, and you'll get an explanation after each one.
Passing this checkpoint: work through the level first, then score 70% or more here.
Skipping this level: already confident? Take this cold and score 80% or more to test out and jump ahead. Below that, nothing is lost, you just study the level as normal.
1. You're running Fernway's office move over several weeks and keep pasting the same lease terms, cost comparison and floor plan into every new chat. What's the practitioner move? Keep pasting them each time; it's only a few files. Create a Project, add those files as project knowledge, and write project instructions, so every chat inside it starts with the context already there. Put all the files and every chat into one endless conversation so nothing is lost. Email the files to Claude so it remembers them for next time. 2. Claude has built an action tracker in the artifact panel and you want to reorder the rows and add a Priority column. What's the cleanest way? Copy the table into the message box, edit it there by hand, and paste it back. Ask Claude to make the changes, or edit the artifact directly in the panel, where it updates in place and keeps the earlier versions. Start a fresh chat and describe the table you want from memory. Leave it; once an artifact is created it can't be changed. 3. You ask Claude for 'total revenue by region' from a sales spreadsheet. It replies instantly with confident figures. Before you quote them, what should you be sure of? Nothing; Claude read the file, so the totals must be right. That Claude actually computed the figures with the analysis tool rather than eyeballing them, and that you told it how to handle blanks and errors in the data. That the answer came back quickly, which means it's simple enough to trust. That you asked politely, so Claude took extra care. 4. You've given Claude a very large document and asked 'does this contract mention a break clause anywhere?' It says no. Why treat that with more caution than a question about a normal-sized file? Large files are always summarised inaccurately, so nothing from them can be trusted. With a very large file Claude may retrieve and read only the relevant-seeming passages rather than every line, so an 'anywhere?' question can miss something; ask a pointed question and have it quote the lines. Claude refuses to answer questions about long contracts for legal reasons. Large files can only be analysed on a paid plan, so free answers are unreliable. 5. You want Claude to always write your internal notes in a plain, warm, filler-free voice, across every chat, not just one. Where does that belong? Retyped at the top of every prompt, every time. In a saved custom style, which shapes the voice across your chats until you change it. In the memory as a confidential fact about a client. Nowhere; tone can't be controlled in Claude. 6. A colleague notices Claude 'knows' their role and preferences in a fresh chat and worries it's stored something it shouldn't. What's the accurate, reassuring answer? Once memory stores something it's permanent and hidden; nothing can be done. Memory is viewable and editable under Settings, they can delete or correct any item, pause or reset it entirely, and use an incognito chat for anything they don't want kept. Memory only works on paid plans, so a free account has nothing stored. They should start a brand-new account to clear it. 7. Claude's Research mode returns a well-structured report on a market question, full of citations. You're about to put its headline claim in a board paper. What's the wise next step? Use it as-is; a cited, structured report has clearly done the checking for you. Open the links behind the load-bearing claims and confirm the sources actually support them, weighing how authoritative each source is. Ask Claude to add more citations so it looks more thorough. Trust the claims with the most footnotes and ignore the rest. 8. You need a single quick fact, the current UK VAT registration threshold, for an email you're writing now. Which is the right tool, and the right habit? Research mode, because deeper is always safer. Web search for the current figure, then open the official source link to confirm it before you use it. Claude's own memory of the figure, since it's a stable number. Whatever Claude says from training, because tax figures rarely change. 9. You've built a new-starter welcome pack as an artifact inside your onboarding Project. It reads beautifully and draws on the real policy file. What separates a finished deliverable from one that just looks finished? Nothing; it's drawn from the project files, so it must be accurate. Verifying the load-bearing facts (office days, core hours, names, roles) against the source files, ideally by having Claude trace each claim to its source line. Making the formatting even nicer so it looks more professional. Publishing it straight away so the team can start using it. 10. Halfway through a Project you realise a standing rule you set ('always assume the South region') was wrong and is quietly skewing every answer. What's the fix? Start an entirely new Project from scratch and re-upload everything. Edit the project instructions to remove or correct the rule; new chats in the Project will follow the updated guidance. Add a note in each individual chat asking Claude to ignore the rule. Delete the project knowledge files, since the problem must be in them. Answered 0 of 10.
Answer every question to see your score