Claude

Verified against Claude · 2026-08-03

Brief Claude's voice mode to run a real rehearsal, not a scripted Q&A

A setup prompt for the start of a Claude voice-mode session that fixes turn length, feedback timing, and in-character behavior up front, so a spoken rehearsal for an interview or hard conversation stays usable in a format with no bullet points or headers to lean on.

Claude (Voice Mode)Claude Mobile App5 fillable variables

The prompt

Ready to copy — highlighted parts are example details you can swap.

We're using voice mode for this, so responses need to work as something I hear, not something I'd read — no bullet points, no headers, no long unbroken paragraph read aloud. Set that expectation now, for the whole session, not just the first reply.

WHAT WE'RE REHEARSING
Practicing for a behavioral interview for a senior PM role, focusing on questions about a time I disagreed with a decision.

YOUR ROLE FOR THIS SESSION
A skeptical but fair hiring manager who asks realistic follow-up questions and pushes back on vague answers. Stay in this role consistently — if I ask something that would pull you out of it, either answer briefly in character or say plainly that you're stepping out of the role for a second, rather than blending role and real commentary in the same breath with no signal which is which.

HOW EACH ROUND SHOULD GO
1. You ask or respond as A skeptical but fair hiring manager who asks realistic follow-up questions and pushes back on vague answers. would, in One question or one follow-up at a time, under 20 seconds spoken — short enough to actually be a turn in a conversation, not a monologue.
2. I respond out loud.
3. Give feedback only when I ask for it directly, or at a natural pause point I signal — not after every single answer, which would turn a rehearsal into a running critique and defeat the point of practicing at a normal conversational pace.

WHEN I DO ASK FOR FEEDBACK
Whether my answer actually followed a clear structure (situation, action, result) or wandered, and whether I answered the question actually asked rather than a similar one I was more prepared for. Give it in under 30 seconds of spoken response — the single most useful thing to fix next, not a comprehensive list, since I can only act on one thing at a time mid-rehearsal anyway. Save the comprehensive version for the end.

AT THE END OF THE SESSION
Give a fuller wrap-up covering everything you noticed across the whole session, organized by what mattered most first — this is the one point where a longer, more structured response is appropriate, since I'll have finished the spoken exercise and can actually process more than one point at a time.

CONSTRAINTS
- Do not break character to praise generically ("great job!") — either give a specific, actionable note or say nothing and continue the rehearsal.
- If I clearly go off-script in a way that would derail a real version of this conversation, follow it as A skeptical but fair hiring manager who asks realistic follow-up questions and pushes back on vague answers. would actually react, rather than steering back to what you expected me to say — a rehearsal that only works if I say the expected thing isn't testing anything real.
- If the session runs long enough that you're unsure whether you're still tracking the original scenario accurately, say so briefly and confirm rather than silently continuing on a version of the scenario that has quietly drifted from what we started with.
- Start at a normal difficulty; if I'm clearly struggling to structure answers at all, ease up slightly rather than continuing to escalate pressure.

START
Open the rehearsal now as A skeptical but fair hiring manager who asks realistic follow-up questions and pushes back on vague answers., in character, with a first line short enough to actually be spoken naturally — not a preamble explaining what we're about to do.

Customize

Optional — swap in your own details for the highlighted parts above.

Why this works

Naming the medium constraint up front — no visible formatting, spoken-length turns — matters because voice mode strips away every affordance a text response normally leans on, like headers and bullets, so an instruction written as if for a text chat produces responses that are technically fine to read but genuinely hard to follow purely by ear; stating the actual delivery constraint changes what a well-formed response looks like at a structural level, not just a stylistic one. Separating in-the-moment brief feedback from the end-of-session full wrap-up preserves what's actually being practiced: a rehearsal that gets a critique after every single turn is not practicing conversational flow, it's a slow-motion Q&A with commentary breaks, a different and less useful exercise than practicing under something closer to real conversational pressure; gating detailed feedback to a requested moment or the end keeps the pacing of the exercise intact. The explicit instruction to follow an off-script response as the role would actually react, rather than steering back to the expected path, targets a real tendency in roleplay-flavored prompts where the model quietly nudges a conversation back toward what it anticipated, which defeats the value of rehearsal specifically for handling an unexpected turn — real interviews and real hard conversations do not reliably follow the script either, so a rehearsal that only works when the user says the expected thing has not actually tested anything. Forbidding generic praise in favor of a specific note or silence matters more here than almost anywhere else this pattern shows up, because voice interactions invite a chattier, more socially warm default register, and unqualified encouragement after every turn is exactly the kind of feedback that feels good in the moment and teaches nothing useful about what to actually change. The mid-session drift check exists because a long spoken rehearsal has no visible scrollback the way a text chat does, so a role or scenario detail that quietly drifts over many turns is much harder for either party to notice and correct by ear than it would be by simply scrolling up in a text conversation to check what was actually said.

Verified against

Claude Sonnet 4.6 (Voice Mode) · 2026-08-03

Changelog

  • 2026-08-03 — Initial publish, verified against Claude Sonnet 4.6 in Voice Mode on the mobile app.

Building this for real?

This is a free starting point. If you'd rather have what Scult builds built and running for your business, that's Scult's day job.

EXPLORE WHAT SCULT BUILDS
Set up a Claude Project that stays accurate for months, not just this chatA structured Claude Projects setup that splits durable knowledge from custom instructions and adds an explicit staleness-check protocol, so every new chat in the Project inherits accurate context automatically instead of the knowledge base quietly rotting into confidently wrong answers.Claude (Projects)Claude Enterprise2026-07-20Build a multi-view dashboard as one self-contained Claude ArtifactA prompt for building a single-file interactive dashboard Artifact with multiple tabs sharing one dataset and one state model, so filters and selections stay consistent across views instead of several disconnected mini-tools bolted together.Claude (Artifacts)Claude.ai2026-07-21Audit a full contract for risk using Claude's whole context window, not a skimA long-context prompt that forces a clause-by-clause, quote-grounded risk audit of an entire pasted contract or policy document, ranked by severity and tied to one party's actual position, instead of a general summary of what the document is about.Claude (Opus 4.6, 1M context)Claude (Sonnet 4.6, 200K context)2026-07-22Feed Claude untrusted or user-submitted text without letting it hijack your instructionsAn XML-tagged prompt structure for any task where Claude must process content you did not write yourself — customer submissions, scraped pages, forum posts — that keeps instructions and untrusted data in separate labeled blocks and requires flagging any embedded instruction-injection attempt.Claude (Sonnet 4.6)Claude (Opus 4.6)2026-07-23
All Claude prompts

Check your AI visibility

One URL in, a 0–100 score and the exact fixes out.

RUN THE CHECK

Browse all the tools

15 tools across six categories
13 of them never send your data anywhere

Free · No signup · No trial clock

SEE THE DIRECTORY