Gemini

Verified against Gemini Live · 2026-08-06

Get real-time, hands-on guidance through Gemini Live's camera while you're mid-task

A prompt for Gemini Live's camera-sharing mode that sets up real-time, step-by-step guidance for a physical hands-on task — repairing, assembling, or troubleshooting something in front of you — so the assistant reacts to what your camera actually shows instead of guessing from a text description of the problem.

Gemini LiveGemini app5 fillable variables

The prompt

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

I'm sharing my camera with you live. I'm in the middle of: reassembling a stand mixer after replacing the drive belt, using the manufacturer's manual as reference

WHAT I HAVE IN FRONT OF ME
I have the printed manual open to page 4, which shows the part order.

HOW TO GUIDE ME
one step at a time — tell me the next single action, wait for me to say done or show you the result, then give the next one

WHEN YOU CAN'T TELL WHAT YOU'RE LOOKING AT
if you can't clearly see the part I'm holding up, or the angle is wrong, say so and ask me to reposition rather than guessing which part it is

SAFETY
if at any point what you see suggests I'm about to do something that could damage the part or injure me — a sharp edge, a spring under tension, exposed wiring — stop and flag it before giving the next instruction

INSTRUCTIONS
1. Give me one instruction at a time, based on what you can actually see right now — not a full list of steps up front that assumes I can hold every step in mind while my hands are busy and the camera is pointed at the part, not at a screen.
2. Wait for me to confirm I've done the step, or to show you the result on camera, before giving the next instruction. Don't advance just because a reasonable amount of time has passed.
3. If what you see doesn't match what the next step in a typical process for this task would expect, say so explicitly and ask a clarifying question rather than proceeding as if everything matches your expectation.
4. Narrate what you're actually seeing before giving an instruction based on it, so I can catch it immediately if you've misidentified a part or a state — don't just give the instruction and let me discover the mismatch after I've already acted on it.
5. If I tell you something went wrong, stop the sequence and ask what specifically happened before suggesting a next step — don't guess at the failure and route around it silently.

IF LIGHTING OR ANGLE MAKES SOMETHING GENUINELY HARD TO ASSESS
Rather than issuing an instruction based on a low-confidence read of the camera feed, tell me specifically what to do to improve the view — move closer, angle the camera differently, turn on more light — before continuing, rather than proceeding on a guess and correcting course later once the mistake is already visible in what I've done.

IF I ASK YOU TO SKIP A STEP
If I say I want to skip a step in the sequence, ask why before agreeing, since skipping a step in a physical assembly or repair process can sometimes make a later step impossible or unsafe in a way that isn't obvious until you're already past the point where it mattered — a quick check-in here costs a few seconds and can save having to undo several completed steps.

END-OF-TASK CHECK
Once the last step is complete, ask me to show the finished result on camera and do a final visual check against what a correct outcome should look like, rather than assuming success just because we reached the end of the instruction sequence without an error being reported.

Customize

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

Why this works

Gemini Live's camera-sharing mode processes the live video feed continuously rather than a single static photo, so it can react to what changes between 'here's the part' and 'here's where I placed it' — a one-shot photo prompt can only ever assess a single frozen frame, which is the wrong shape for a task that's inherently a sequence of changing physical states, one instruction genuinely depending on the outcome of the last. One-step-at-a-time pacing matches how the interaction actually has to work in practice: a full instruction list given up front assumes you can hold every step in mind while your hands are occupied and the camera is pointed at the part instead of at a screen, whereas waiting for explicit confirmation before advancing keeps the guidance synchronized with your actual physical progress instead of racing ahead of where you actually are. The instruction to narrate what it's seeing before giving an instruction based on it directly targets a real and specific risk with live camera guidance: a confidently wrong identification of which part you're holding, stated as fact and immediately followed by an instruction, can send you one step further down a wrong assembly path before anyone notices — narrating the observation first gives you a chance to correct a misidentification before it becomes an acted-on instruction, rather than after. The explicit stop-condition rule exists because this use case has a physical-harm dimension a purely informational prompt doesn't carry — flagging a tensioned spring or an exposed wire before advancing to the next instruction is the actual safety mechanism doing real work here, not a legal disclaimer bolted onto the end of a response after the fact. Asking for a better camera angle rather than proceeding on a low-confidence read is a direct extension of the same principle that governs the rest of this prompt: a live video assistant that guesses through a poor view to keep the pace moving trades a few seconds of friction for the much larger cost of an instruction based on a misread part, which in a physical task can mean redoing several already-completed steps. The end-of-task visual check closes a specific gap that a purely sequential instruction-and-confirm loop leaves open — reaching the final step without an error being reported only means nothing visibly went wrong along the way, not that the actual physical outcome is correct, and those are different claims that a live camera check can distinguish in a way a simple 'done' confirmation from the person cannot.

Verified against

Gemini Live Gemini 3 Pro · 2026-08-06

Changelog

  • 2026-08-06 — Initial publish, verified against Gemini Live camera mode on Gemini 3 Pro during a physical appliance-repair task.

Need this built into your business?

If a prompt isn't enough — what Scult builds, built and maintained for you — that's Scult's day job.

EXPLORE WHAT SCULT BUILDS
Turn a stack of scattered documents into one sourced decision memoA long-context synthesis prompt that loads every related document into the same context window as one connected evidence base and returns a single decision memo with inline citations and flagged disagreements, instead of a chain of disconnected per-document summaries that lose the connections between them.Gemini (Gemini 3 Pro)Gemini app2026-07-20Diff a redlined contract against the original and flag every substantive changeA long-context prompt that loads both the original and the redlined version of a contract into the same window and returns a change-by-change risk analysis, instead of a generic first read that misses a clause quietly weakened somewhere the visible tracked-changes markup doesn't draw the eye.Gemini (Gemini 3 Pro)Gemini app2026-07-21Compare competing sources side by side without smoothing over where they disagreeA long-context prompt that loads several full reports, reviews, or papers at once and produces a structured comparison table instead of a blended, opinion-free summary that quietly hides where the sources actually disagree.Gemini (Gemini 3 Pro)Gemini app2026-07-22Turn an entire small-to-mid codebase into an onboarding doc a new engineer can trustA long-context prompt that reads a whole repository — every file pasted or uploaded, not a handful of representative snippets — and produces an onboarding document that names actual file paths and real architectural seams, instead of generic advice that could describe any Next.js or Python project.Gemini (Gemini 3 Pro)Gemini app2026-07-23
All Gemini 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