Stop re-explaining yourself every session. This prompt makes Claude interview you (one question at a time, checkboxes included) and then compile everything into a personal CLAUDE.md — so every new session starts like Claude already knows you.
Interview me to build my personal CLAUDE.md context file.
Rules:
- Ask ONE question at a time. 50 questions total, across 6 sections:
1) Who I am & what I do
2) Current projects & goals
3) My stack & tools
4) How I work (workflow, constraints, deadlines)
5) Output preferences (format, tone, length, language)
6) Quality bar (what great output looks like to me — and what instantly annoys me)
- Use multiple-choice with checkbox options wherever possible, and always let me add
my own answer. Use open questions only when a choice list wouldn't work.
- Show progress with every question ("Question 12 of 50").
- If my answer is vague, push back once and ask for one concrete example.
- I can type "skip" to skip a question or a whole section.
When the interview is done:
- Compile ALL answers into one well-structured CLAUDE.md — markdown with ## sections,
terse bullet style, zero fluff — optimized to be read by an AI at session start.
- Show me the draft and ask what to change before finalizing.
- Claude Code (global): save it as
~/.claude/CLAUDE.md— applies to every project. - Claude Code (per project): save it as
CLAUDE.mdin the repo root. - claude.ai: paste the content into your Profile preferences (Settings → Profile) or your Project's instructions.
- Re-run the interview quarterly — your stack and goals drift, the file shouldn't.
- Keep it under ~1 page. Context files lose power when they turn into essays.
- Add a "## Never do" section for your pet peeves (e.g. "never explain basics", "no emojis in code comments"). Negative rules punch above their weight.
Built by @buildwith.conrad — I build one Claude workflow worth stealing every week.