Test yourself on this chapter. Try each question before reading the answer.
1. What's the biggest quality lever when asking Claude to analyze real customer feedback?
- A. Asking it to be creative
- B. Telling it not to invent anything beyond the pasted data
- C. Keeping the prompt as short as possible
- D. Asking for bullet points
Answer: B. Grounding constraints ("only use my notes, don't invent") keep the output anchored to real evidence: the #1 fix for generic AI answers.
2. In the 5-part prompt skeleton, which part is 'a table with columns Theme | Evidence | Severity'?
- A. Role
- B. Context
- C. Format
- D. Constraints
Answer: C. Naming the exact sections or columns you want back is the Format part: it's what makes output copy-paste ready.
3. After Claude drafts your brief, what's the recommended next move?
- A. Ship it to engineering as-is
- B. Ask Claude to make it longer
- C. Ask Claude to attack its weakest assumptions, then fix them yourself
- D. Start over with a different prompt
Answer: C. Draft → critique → human judgment is the core loop. Claude drafts and attacks; you decide.