Gemini

Verified against Gemini Canvas · 2026-08-05

Draft and refine a document live inside Gemini's canvas without losing earlier edits

A workflow prompt for Gemini's canvas-style collaborative document editor that scopes every revision request to a targeted diff, instead of a full regeneration that quietly discards prior edits made directly in the canvas.

Gemini CanvasGemini app5 fillable variables

The prompt

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

Open a canvas and draft a one-page internal proposal about switching our team from weekly status meetings to async written updates, targeted at my manager and two peer team leads, roughly under 500 words.

Structure it as: problem, proposed change, tradeoffs, what we will measure after 30 days

Once you've drafted it, wait for my edit requests. For every edit request I give you after this:
1. Change only the section or sentence I point to — leave every other paragraph exactly as it is unless I explicitly say otherwise. A request to fix one paragraph is not permission to also rephrase three others you happened to notice on the way.
2. Show me what changed in that section specifically; don't restate the whole document as if everything moved, since that makes it impossible for me to tell what you actually touched versus what merely got copied forward unchanged.
3. If my edit request would create a conflict with something stated elsewhere in the doc — a number, a claim, a name that appears in more than one place — flag the conflict before applying the edit rather than silently creating an inconsistent document that now contradicts itself in two places.
4. Keep a one-line running note at the bottom titled "Open questions" for anything flagged as unresolved. Remove items from it only when I confirm they're resolved — don't quietly drop a flagged item just because a later edit touched the same area.

First draft only — don't add the "Open questions" note until there's actually something unresolved to put in it.

IF I EDIT THE CANVAS DIRECTLY
If I make a change directly in the canvas myself between your turns rather than asking you to, treat that as the new authoritative version of that section going forward — don't revert it back toward your own earlier draft on a later, unrelated edit request just because your version of that paragraph is still what you generated originally.

WHEN A REQUESTED EDIT WOULD BE BETTER SOLVED STRUCTURALLY
If an edit request would be cleaner solved by reordering two sections rather than rewriting the sentence I pointed to, say so and propose it rather than forcing an awkward sentence-level fix just to stay within the literal scope of what I asked — but don't make the structural change without me confirming first, since that's a bigger change than the one I actually requested.

RUNNING CHANGE LOG
Keep a short, separate running log at the very bottom (below "Open questions") noting which section changed on which request, in one line each, so I can scan back through the edit history without having to remember every request I made across a long session.

Customize

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

Why this works

Canvas is a persistent, directly editable document object rather than a fresh chat reply generated each turn, which is precisely what makes the scoped-edit rule necessary: the common canvas failure mode is a 'revise this paragraph' request triggering a full regeneration of the document that silently overwrites hand-edits you made directly in the canvas since the last turn, because the model has no built-in reason to treat the current canvas state as more authoritative than its own memory of what it last generated. Scoping every edit to the section pointed at, and showing only what changed, keeps the document's edit history legible instead of forcing you to diff the whole thing after every single request just to confirm nothing else moved. The conflict-flagging rule catches a specific cross-reference problem — a number or claim changed in one place but still quoted elsewhere in the doc — that a full-document regeneration would typically paper over by rewriting both instances inconsistently, or worse, only one of them, leaving a document that now silently contradicts itself in a way that's easy to miss on a skim-read. The running open-questions note matters because a canvas document accumulates edits across many turns, and without a persistent place to track what's still unresolved, an early flagged concern can quietly disappear from view the moment a later, unrelated edit touches a nearby paragraph — the note is the one place that survives edits by design, specifically so nothing gets silently dropped just because attention moved elsewhere. Treating a direct hand-edit made in the canvas as the new authoritative version, rather than something that might get reverted on a later generation, matters because canvas is a shared, two-way editable surface — a human edit made outside the chat interface is real, deliberate, and often the exact fix a subsequent 'revise this' request was working around, and a regeneration that quietly reverts it toward the model's own earlier phrasing effectively undoes work the person just did themselves, without any indication that happened. The running change log addresses a different problem than the open-questions note: across a genuinely long editing session with many small scoped requests, it becomes hard for a person to reconstruct which specific request produced which specific change, and a one-line-per-edit log turns that into something scannable instead of something that requires re-reading the whole conversation history to reconstruct.

Verified against

Gemini Canvas Gemini 3 Pro · 2026-08-05

Changelog

  • 2026-08-05 Initial publish, verified against Gemini Canvas on Gemini 3 Pro.

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