Claude AI Master Quiz
Ten questions spanning everything Claude — prompting, coding, writing, research, workflows, hidden features, and the mistakes that quietly waste your turns. Take it as a diagnostic: whatever you miss tells you which topic to read next. No sign-up, instant score, and you can share how you did.
-
Q1 You open a fresh chat and ask Claude to 'write a follow-up email'. Why is the answer generic?Why
Claude has broad general knowledge and zero knowledge of your circumstances. Stripped of the background only you have, it must invent one — and invented context produces generic output.
-
Q2 Which is the correct mental model of Claude's context window?Why
It's working memory, not storage. Everything shares one budget — which is why a huge paste degrades quality and why a new chat remembers nothing unless you re-supply it.
-
Q3 You want Claude to fix a bug. What gives you the best shot at a real fix?Why
A reproduction pins Claude to a real theory of the failure. Without a failing case it guesses a fix that looks right but doesn't address the cause.
-
Q4 Before Claude drafts a long article, what's the highest-leverage step?Why
Reordering an outline bullet is free; restructuring four drafted sections is a rewrite. Approve the structure before any prose exists.
-
Q5 You want a summary whose claims you can check. What do you ask for?Why
Quotes make claims traceable rather than something you take on trust. Spot-check them — an unverified citation is decoration — and keep what the source states separate from what Claude inferred.
-
Q6 What should one step of a prompt chain hand to the next?Why
Piping prose between steps causes silent field drops — something goes missing and nothing errors. Structure makes the handoff checkable.
-
Q7 What is MCP (the Model Context Protocol)?Why
MCP turns Claude from a chat window into something that can act on your real data. Scope each server to least privilege — read-only when you're only reading.
-
Q8 What's the test for whether something belongs in your context?Why
Completeness is not relevance. Anything that wouldn't change a correct answer is diluting the signal — and stale material makes Claude reconcile conflicts you never meant to introduce.
-
Q9 You ask 'This is the better option, right?' and Claude agrees. What have you learned?Why
A loaded question gets a loaded answer. Treating that agreement as an independent second opinion is the trap — the frame already decided the response.
-
Q10 Claude returns code that's well-structured and reads correctly. Ship it?Why
Presentation is not proof, and asking the model 'are you sure?' rarely surfaces the flaw. A real check does — and the happy path is exactly where AI code doesn't fail.