Claude Prompt Library

35 Claude Prompts for Google Docs

35 copy-paste prompts

Claude has no button inside Google Docs, so these 35 prompts work the way the job actually gets done: you paste the doc, the comment threads, or the transcript into Claude, and it returns the draft, the exact find and replace list, or the Apps Script you paste into Extensions.

In short: This page contains 35 copy-paste ready prompts, organized into 5 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Hand-curated and tested by the AI Academy team.

Louis Corneloup
By Louis Corneloup · Founder, Techpresso
Last updated ·Hand-curated & tested by the AI Academy team

Drafting & Outlines

7 prompts

Outline From a Brain Dump

1/35

✨ What it does

Claude turns your raw notes into a Google Docs outline with one Heading 1, then Heading 2 and Heading 3, aimed at [audience] and [main question]. You paste the outline into the doc and write the body under those headings.

<context>[Paste your raw notes, bullets, and half finished sentences for the document.] The reader is [audience] and the doc has to answer [main question].</context> <task>Turn this into a Google Docs outline: one Heading 1, then Heading 2 sections, then Heading 3 subpoints where the content justifies them.</task> <format>Outline only, one line per heading, with the heading level written in front of each line, plus a single line under each H2 naming the evidence that section still needs.</format>

💡

Pro tip: Paste the outline in, then select each line and apply the matching paragraph style. The outline pane on the left fills in immediately and becomes your progress tracker.

First Draft From an Approved Outline

2/35

✨ What it does

Claude writes a complete first draft of about [N] words that follows the approved outline and adds nothing outside it. You paste the draft into Google Docs and edit the voice before anyone reviews it.

<context>[Paste the approved outline with its heading levels.] Target length is [N] words, the audience is [audience], and the doc will be read [live in a meeting / async before a decision].</context> <task>Write the full first draft, keeping the exact heading structure and adding nothing outside it.</task> <format>Plain text with headings on their own lines, paragraphs of three sentences maximum, and any figure I have not supplied written as [TBD: what you need from me].</format>

💡

Pro tip: The [TBD] markers are the point. Ctrl+F for 'TBD' in Docs and you have an instant list of every fact you still owe the reader.

Decision Doc From Meeting Notes

3/35

✨ What it does

Claude turns raw meeting notes about [topic] into a decision record: what was settled, options considered, who owns next steps, and what would reopen it. You paste it into Google Docs so people who missed the meeting can still follow it months later.

<context>[Paste the raw meeting notes or transcript.] The meeting was about [topic], and the people who were not in the room need to understand what was settled.</context> <task>Write a decision doc: the decision, the context behind it, the options considered, why the others lost, who owns the next steps, and what would make us revisit it.</task> <format>Six short sections with those exact headings. Anything discussed but not decided goes in a final Open Questions list, with the name of whoever raised it.</format>

💡

Pro tip: Write the Open Questions list while names are still fresh. That list is what stops the same debate resurfacing in six weeks.

Project Brief From a Kickoff Transcript

4/35

✨ What it does

Claude drafts a project brief for [name] covering problem, goal, non-goals, approach, stated dates, owners, risks, and dependencies. You check which claims were inferred, then send the brief to [stakeholders].

<context>[Paste the kickoff call transcript or notes.] The project is [name] and the brief goes to [stakeholders].</context> <task>Draft a project brief covering problem, goal, explicit non goals, approach, milestones with the dates that were actually stated, owners, risks, and dependencies.</task> <format>Those headings in that order. For every claim, mark whether it was stated in the transcript or inferred by you, so I only have to verify the inferred ones.</format>

💡

Pro tip: The stated versus inferred split stops you shipping a brief where half the scope was invented. Check every inferred line with its owner before you share the link.

SOP Draft From Scattered Instructions

5/35

✨ What it does

Claude writes a numbered standard operating procedure for [name] that [role] can follow without asking a single question. You paste it into Google Docs and walk through it once on a real run.

<context>[Paste every instruction you have for this process: chat messages, old emails, notes, whatever exists.] The process is [name], it runs [frequency], and it is performed by [role].</context> <task>Write a standard operating procedure a new hire could follow without asking a single question.</task> <format>Numbered steps, one action per step, the tool used in bold at the start of each step, a Prerequisites block at the top and a What Goes Wrong block at the bottom.</format>

💡

Pro tip: One action per step is the rule that makes an SOP usable. If a step contains the word 'and', split it before you publish.

Section Word Budget Plan

6/35

✨ What it does

Claude assigns a word budget to each section so the parts that drive the decision get more of the [5 / 10 / 20] minutes the reader will give. You trim or expand each section to those numbers.

<context>My doc is [current length] words and the reader will give it [5 / 10 / 20] minutes. Here is the outline: [paste].</context> <task>Assign a word budget to each section based on how much that section contributes to the decision the reader has to make.</task> <format>A table of section, word budget, and one line justifying the budget. Flag any section whose budget lands under 100 words as a candidate to cut entirely.</format>

💡

Pro tip: Keep the table pinned at the top of the doc while you draft, then delete it before sharing. Select any section and use Tools then Word count to check yourself against it.

Reusable Doc Template With Bracketed Fields

7/35

✨ What it does

Claude pulls the strongest structure from your two recent [doc type] examples and writes a blank reusable template with bracketed fields. You save it as the team template and fill those fields each time.

<context>My team writes [doc type] roughly [frequency] and every one comes out different. Here are two recent examples: [paste both].</context> <task>Extract the strongest structure from the two and write a blank reusable template from it.</task> <format>The template with every variable part as a [BRACKETED FIELD], plus one instruction line under each heading saying what belongs there and, just as important, what does not.</format>

💡

Pro tip: Save it in a Drive folder as a view only original and tell the team to use File then Make a copy. Nobody edits the master that way.

XML tags are just the start. Learn the full Claude workflow.

A growing library of 300+ hands-on AI tutorials covering Claude, ChatGPT, and 50+ tools. New tutorials added every week.

Start 7-Day Free Trial

Editing & Rewriting

7 prompts

Cut the Doc to Length Without Losing Facts

8/35

✨ What it does

Claude trims the doc from [current] words to under [target] while keeping every number, name, date, and commitment, and shows what was dropped. You review the cut list, then paste the shorter version into Google Docs.

<context>[Paste the full doc.] It is [current] words and has to come in under [target] words. Every number, name, date, and commitment has to survive the cut.</context> <task>Trim it. Remove throat clearing, repeated context, and any sentence that only restates the heading above it.</task> <format>The trimmed doc first, then a list of what you removed grouped by reason, so I can restore anything you judged wrong.</format>

💡

Pro tip: Read the removal list before the draft. It usually reveals that two of your sections were making the same point in different words.

Plain Language Rewrite for a Non-Expert Reader

9/35

✨ What it does

Claude rewrites the doc so [new audience, for example the finance team or a customer] understands it on one read, without stripping the substance. You keep the same headings and check any term that still feels off.

<context>[Paste the doc.] It was written for [original audience] and now has to be read by [new audience, for example the finance team or a customer].</context> <task>Rewrite it so the new audience gets it on one read, without stripping out the substance.</task> <format>Same headings. Replace jargon with plain wording, define any term that has to stay in a parenthesis at first use, and end with a list of every term you kept and why.</format>

💡

Pro tip: The kept-terms list is your glossary. Paste it at the bottom of the doc as an appendix instead of deleting it.

Tone Match to Your Team's Existing Docs

10/35

✨ What it does

Claude rewrites your new draft so it reads like the two team documents people already responded well to. You compare the style note to those originals, then paste the rewrite if the voice matches.

<context>Here are two documents my team wrote that people responded well to: [paste both]. Here is my new draft: [paste].</context> <task>Rewrite my draft so it reads like it came from the same team.</task> <format>The rewritten draft, then a style note listing the specific patterns you matched: sentence length, how we open sections, how we deliver bad news, words we use, words we avoid.</format>

💡

Pro tip: Keep the style note. Paste it as context in future prompts and you can skip the sample documents entirely from then on.

Passive Voice and Hedging Sweep

11/35

✨ What it does

Claude finds every passive line and hedge in the [doc type] and rewrites each with a named actor and a direct verb. You apply the rows you agree with and leave the ones that should stay cautious.

<context>[Paste the doc.] It is a [doc type] where somebody has to be accountable for each action described.</context> <task>Find every passive construction and every hedge such as 'we may want to consider', and rewrite them with a named actor and a direct verb.</task> <format>A table of original sentence, rewrite, and the actor you assigned. Where you cannot tell who the actor is, write UNKNOWN OWNER rather than guessing a name.</format>

💡

Pro tip: The UNKNOWN OWNER rows are the real output. Each one is work that nobody in the room has agreed to do.

Executive Summary That Survives a Skim

12/35

✨ What it does

Claude writes an executive summary for [role] that stands alone: the ask, the recommendation, three reasons, the cost, and the main risk, plus a chat-sized version. You put the short one at the top of the Google Doc.

<context>[Paste the full doc.] The reader is [role] and will read the first screen only before deciding whether to keep going.</context> <task>Write an executive summary that stands completely alone: the ask, the recommendation, the three reasons, the cost, and the main risk.</task> <format>Under 150 words, no bullet longer than two lines, the ask in the first sentence. Then a second version at half that length for a chat message.</format>

💡

Pro tip: Put the summary above the table of contents, not below it. Readers stop at the first thing that looks like a wall of text.

Reverse Outline: What the Doc Actually Says

13/35

✨ What it does

Claude reverse outlines the doc, ignoring [your intended argument]: for each section, the claim it actually makes and the evidence it actually gives. You rewrite the sections that do not match what you meant.

<context>[Paste the doc.] I think it argues [your intended argument], but I have read it too many times to judge.</context> <task>Ignore my stated intent and reverse outline it: for each section, write the one claim it actually makes and the evidence it actually provides.</task> <format>A numbered list in section order, then a verdict on where the real argument drifts from my intended one and which section is doing the least work.</format>

💡

Pro tip: Run this before the review cycle, not after. Reviewers who spot a structural drift will never bother commenting on your wording.

Consistency Pass on Names, Terms and Capitalization

14/35

✨ What it does

Claude builds a find-and-replace list for terminology, capitalization, product names, dates, and numbers that break house style. You run those replacements in Google Docs instead of accepting a full rewrite.

<context>[Paste the doc.] House style: product names are [list], we write [term] and never [variant], dates use [format], numbers over [threshold] use [convention].</context> <task>Find every inconsistency in terminology, capitalization, product naming, date format, and number format.</task> <format>A find and replace table with three columns: exact current text, exact replacement text, number of occurrences. Do not return a rewritten document.</format>

💡

Pro tip: Docs has find and replace under Edit, with match case and regex options. Working from an exact-text table is faster and far safer than pasting a whole rewritten doc back over your own.

Collaboration & Comments

7 prompts

Comment Threads to an Action List

15/35

✨ What it does

Claude turns every comment thread on [name] into something you can act on: fact fix, opinion, or scope change, plus what to do before [date]. You work the list in that order and close the review.

<context>[Paste the comment threads, including who wrote what and any replies.] The doc is [name] and I have to close the review by [date].</context> <task>Turn every thread into something I can act on: what is being asked, whether it is a fact fix, an opinion, or a scope change, and what I should do about it.</task> <format>A table of commenter, thread summary, type, recommended action, and whether it can be resolved without a reply. Sort it so the fact fixes come first.</format>

💡

Pro tip: Open the comment history icon at the top right of the doc to copy every thread at once, including the ones already resolved, instead of scrolling the sidebar.

Reply Drafts for Every Open Comment

16/35

✨ What it does

Claude drafts one reply under 40 words for each open comment: accept, accept with a change, or decline with a reason. You paste each reply onto the comment from your [peers / manager / client] and send.

<context>[Paste the open comments together with the paragraphs they are attached to.] These reviewers are my [peers / manager / client], and I disagree with [which ones].</context> <task>Draft one reply per comment: accept, accept with a modification, or decline with a reason.</task> <format>Under 40 words per reply, no apologies, and for every decline put the reason before the decline so it does not read as a brush off.</format>

💡

Pro tip: Reply then resolve, never resolve silently. A thread that gets closed with no answer is the fastest way to lose that reviewer next time.

Merge Two Conflicting Versions

17/35

✨ What it does

Claude produces one merged document from Version A and Version B, and lists genuine contradictions separately from wording differences. You decide those contradictions, then paste the merge into Google Docs.

<context>Version A: [paste]. Version B: [paste]. They were edited in parallel by different people and both contain changes worth keeping.</context> <task>Produce one merged document, and separate genuine contradictions from places where the two versions merely word the same thing differently.</task> <format>The merged doc first. Then a conflicts list with the passage, what A says, what B says, and which one you kept, so the two authors can settle the real disagreements themselves.</format>

💡

Pro tip: Pull both versions from File then Version history, and name the versions before you start merging so you can still find them afterwards.

Suggestion Mode Edit List

18/35

✨ What it does

Claude proposes edits as exact original text, exact replacement, and a one-line reason, instead of returning a rewritten document. You apply the suggestions one at a time in suggestion mode.

<context>[Paste the doc.] I am editing somebody else's document and I want to leave suggestions rather than silent rewrites.</context> <task>Propose your edits as exact replacements instead of returning a rewritten document.</task> <format>A table of exact original text, exact replacement text, and a one line rationale I can paste as the suggestion comment. Keep each original snippet short enough to select in one drag.</format>

💡

Pro tip: Switch to Suggesting mode from the pencil icon at the top right before you apply any of these. In Suggesting mode the author keeps the accept or reject decision, which is the whole point.

Reviewer Brief

19/35

✨ What it does

Claude writes a short review request for each of [names and roles], telling them which sections to read and what kind of feedback you want by [date]. You send those messages with the Google Docs link.

<context>[Paste the doc or its outline.] The reviewers are [names and roles] and I need their input back by [date].</context> <task>Write the request that goes out with the review, telling each reviewer which sections to read and what kind of feedback I want from them specifically.</task> <format>A short message per reviewer naming their sections, the question they should answer, and what to ignore. Then one line I can paste at the top of the doc stating the review deadline.</format>

💡

Pro tip: Use an @mention with a date in the doc so each reviewer gets an assigned action item in Drive, not just an email notification they can archive.

Vague Feedback to Specific Edits

20/35

✨ What it does

Claude translates each vague comment into two or three concrete edits and names the passage that probably triggered it. You apply the edits that fit, or ask the reviewer the leftover questions.

<context>Feedback received: [paste it, for example 'this feels too long' or 'not sure about the framing']. Here is the doc it refers to: [paste].</context> <task>Translate each vague comment into two or three concrete edits that would satisfy it, and name the passage that probably triggered it.</task> <format>Grouped by comment: your interpretation, the likely trigger passage, the specific edits, and one clarifying question to send back if your interpretation is a guess.</format>

💡

Pro tip: Send the clarifying question before doing the edits. Half the time the reviewer meant something you would never have guessed from the comment.

Version Diff Explained in Plain English

21/35

✨ What it does

Claude explains what changed in meaning between the previous and current versions: what was added, removed, or moved, and what a reader of the old version would miss. You send that summary instead of asking people to reread.

<context>Previous version: [paste]. Current version: [paste]. Stakeholders have already read the previous one and will not read the whole thing again.</context> <task>Explain what changed in terms of meaning rather than wording: what was added, what was removed, which commitment moved, and what a reader of the old version would now get wrong.</task> <format>Four short lists with those labels, then a two sentence change note I can send to everyone who read the old version.</format>

💡

Pro tip: Version history shows what changed but never why it matters. Name the new version in Docs and paste the change note at the top so both live together.

These prompts give you the what. Tutorials give you the why.

Learn when to use extended thinking, how to build Claude Projects, and workflows that compound. 300+ tutorials and growing.

Try AI Academy Free

Formatting & Structure

7 prompts

Heading Hierarchy Fix

22/35

✨ What it does

Claude rebuilds heading levels, makes the phrasing parallel, and rewrites each heading so it states what the section concludes. You apply the new hierarchy so the Docs outline pane reads like a summary.

<context>[Paste the doc with its current headings marked by level.] People navigate it from the Docs outline pane and keep getting lost in the middle.</context> <task>Rebuild the heading hierarchy: correct the levels, make the phrasing parallel, and rewrite each heading so it states what the section concludes rather than what topic it covers.</task> <format>A before and after list of every heading with its level, plus a note on any section that has no heading and needs one, or has one and should be merged upward.</format>

💡

Pro tip: Apply them with Format then Paragraph styles so the outline pane and the table of contents both update. Bold text at a larger size is not a heading and Docs will not treat it as one.

Wall of Text to Tables and Lists

23/35

✨ What it does

Claude decides what belongs in a table, a numbered list, bullets, or prose, then rewrites the dense section that way. You paste the restructured section and confirm nothing was dropped.

<context>[Paste the section that reads as one dense block.] It contains [comparisons / sequential steps / criteria / options].</context> <task>Decide what belongs in a table, what belongs in a numbered list, what belongs in bullets, and what has to stay as prose. Then rewrite the section that way.</task> <format>Give tables as tab separated rows I can paste directly, and add one line per element explaining why that format suits that content.</format>

💡

Pro tip: Docs converts tab separated text into a real table on paste, so ask for tabs rather than pipes and skip the manual cell entry entirely.

Table of Contents and Link Plan

24/35

✨ What it does

Claude plans navigation for the [N] page doc: which headings belong in the table of contents, which need a shareable link, and what each link should be called. You add those links in Google Docs before you share the URL.

<context>[Paste the heading list of a doc that runs to [N] pages.] It is shared as a link and readers jump straight into the middle of it.</context> <task>Plan the navigation: which headings belong in the table of contents, which sections need a direct link people can share, and what each link should be called.</task> <format>A table of heading, include in the table of contents yes or no, suggested link text, and where else in the doc that link should appear inline.</format>

💡

Pro tip: Insert then Table of contents builds itself from your heading styles, and right clicking any heading gives you Copy heading link for a URL that lands on that exact section.

Split a Long Doc Into a Linked Set

25/35

✨ What it does

Claude proposes a split of the [N] word doc into separate files plus one parent hub, based on who reads what and how often each part changes. You create those docs and move the content.

<context>[Paste the outline of the doc.] It has grown to [N] words and different readers only ever need different parts of it.</context> <task>Propose a split into separate documents with one parent hub doc, based on who reads what and how often each part changes.</task> <format>Per proposed doc: name, owner, contents, audience, update frequency, and which other docs it links to. Then the structure of the hub doc that ties them together.</format>

💡

Pro tip: Split on update frequency, not only on topic. The section you rewrite every month should never sit inside the doc you want to leave frozen for a year.

Process Doc to Numbered Checklist

26/35

✨ What it does

Claude converts the prose process into a numbered checklist [role] can tick through while doing the work, with every decision point written out. You paste it beside the work and try it on the next run.

<context>[Paste the process section, currently written as prose.] It is followed by [role], usually under time pressure.</context> <task>Convert it into a checklist someone can tick through while doing the work, with every decision point made explicit.</task> <format>Numbered steps in checklist form, branch points written as If [condition], go to step [n], and a Verify block at the end listing how to confirm the work was done correctly.</format>

💡

Pro tip: Format then Bullets and numbering then Checklist gives you real checkboxes, and everyone who makes a copy of the doc gets their own independent tick state.

Accessible Document Pass

27/35

✨ What it does

Claude audits heading order, descriptive link text, image alt text, table headers, and meaning carried by color alone, then writes usable alt text. You apply the fixes before the doc is shared outside the team.

<context>[Paste the doc, and describe every image in it: what it shows and why it is there.] It will be shared outside my team.</context> <task>Audit it for accessibility: heading order, descriptive link text instead of 'click here', alt text for each image, table header rows, and any meaning carried by color alone.</task> <format>A findings table of issue, location, and the exact replacement text. Write the alt text for me rather than telling me to write it.</format>

💡

Pro tip: Right click an image and choose Alt text to paste each one in. Screen readers announce link text with no surrounding context, which is exactly why 'click here' fails.

Citations, Footnotes and Appendix Order

28/35

✨ What it does

Claude places each source as an inline link, a footnote, or an appendix item, and flags claims that are not credible enough to cite. You add the citations in [style, or state that there is none] and cut the unsupported lines.

<context>[Paste the doc and the list of sources with their URLs.] House reference style is [style, or state that there is none].</context> <task>Place each reference properly: what becomes an inline link, what becomes a footnote, what moves to an appendix, and what is not credible enough to cite at all.</task> <format>A table of claim, source, placement decision, and the exact text to insert. Separately, flag every claim in the doc that currently has nothing behind it.</format>

💡

Pro tip: Insert then Footnote, or Ctrl+Alt+F, adds footnotes without breaking your numbering. Fix the unsourced claims first, since those are the rows that get challenged in review.

Apps Script for Docs

7 prompts

Find and Replace Across a Whole Drive Folder

29/35

✨ What it does

Claude writes a Google Apps Script that opens every Doc in a Drive folder and applies your replacement pairs, then tells you exactly where to paste and run it. You run it on a copy of the folder first.

<context>I have [N] Google Docs sitting in one Drive folder and need the same replacements applied to all of them: [list the old text and new text pairs]. I have never written Apps Script.</context> <task>Write a Google Apps Script that opens every Doc in a folder and applies the replacements, then tell me exactly where to paste it and how to run it.</task> <format>The full script with comments, the folder ID as a variable at the top, a DRY_RUN flag that only logs what it would change, and numbered run instructions including which permissions it will request.</format>

💡

Pro tip: Run it with DRY_RUN on first and read the execution log line by line. Script edits across files cannot be undone with Ctrl+Z, and version history is per document.

House Style Formatting Normalizer

30/35

✨ What it does

Claude writes a document-bound Apps Script that applies house [font, size, line spacing, heading sizes, margins] to body text and each heading level, without touching tables or images. You bind it to one doc and run it before you share it.

<context>Docs come back to me from different people with mixed fonts, sizes, and spacing. House style is [font, size, line spacing, heading sizes, margins].</context> <task>Write an Apps Script bound to a document that applies the house style to body text and to each heading level, without touching tables or images.</task> <format>The full script with one function per style rule so I can comment out the ones I do not want, plus a short list of what it cannot fix and I will have to do by hand.</format>

💡

Pro tip: Open it from Extensions then Apps Script inside the doc itself so it is bound to that document. A bound script needs no file ID and runs against whatever doc it lives in.

Generate One Doc Per Spreadsheet Row

31/35

✨ What it does

Claude writes an Apps Script that copies the Google Doc template for each Sheet row, replaces placeholders like {{name}} and {{amount}}, and saves each file into [folder]. You run it on a few rows, then the full sheet.

<context>I have a Google Sheet with the columns [list them] and a Google Doc template containing placeholders like {{name}} and {{amount}}. I need one document per row, saved into [folder].</context> <task>Write an Apps Script that copies the template for each row, replaces the placeholders, and names each file from [which columns].</task> <format>Full script with template ID and folder ID as variables, a status written back to the sheet so processed rows are skipped on a re-run, and instructions for launching it from the Sheet.</format>

💡

Pro tip: The status column is what makes the script safe to re-run. Without it, a run that fails halfway creates a full set of duplicates the second time.

Export Doc Comments to a Sheet

32/35

✨ What it does

Claude writes an Apps Script that pulls comments and replies from [IDs, or say all docs in one folder] into a Google Sheet you can filter and assign. You run it, then work the rows like a review tracker.

<context>I run document reviews in Google Docs and need the comments tracked outside the doc. The documents involved are [IDs, or say all docs in one folder].</context> <task>Write an Apps Script that uses the Drive API to pull comments and their replies out of those documents and write them into a Google Sheet.</task> <format>Full script, the exact advanced service I must enable and how to enable it, columns for doc name, author, comment, quoted text, resolved status and date, plus what the API cannot retrieve.</format>

💡

Pro tip: Advanced services are switched on from the Services panel in the Apps Script editor. If the script errors on an unknown identifier, that panel is almost always the reason.

Custom Menu of Team Doc Actions

33/35

✨ What it does

Claude writes an Apps Script with an onOpen custom menu for the actions your team repeats in every document. You install it, click each menu item once, and keep the ones that work.

<context>My team repeats the same actions in every document: [list them, for example insert the standard header, stamp today's date, apply the review checklist, delete empty paragraphs].</context> <task>Write an Apps Script with an onOpen custom menu that exposes each of those actions as its own menu item.</task> <format>Full script with one clearly named function per action, the onOpen menu builder, and an explanation of why the menu only appears after a reload the first time.</format>

💡

Pro tip: The menu only shows for people with edit access who have authorized the script once. Ship it inside a template doc so each teammate authorizes it the first time they make a copy.

Self-Updating Header and Metadata Stamp

34/35

✨ What it does

Claude writes an Apps Script that inserts or updates a top block with [what it should hold, for example doc owner, status, last reviewed date, confidentiality label] and stamps the last modified date. You run it on a test doc, then add it to the team's files.

<context>Our documents need a consistent block at the top containing [what it should hold, for example doc owner, status, last reviewed date, confidentiality label].</context> <task>Write an Apps Script that inserts that block, or updates it if it is already there, and stamps the last modified date pulled from the file metadata.</task> <format>Full script that locates and updates the existing block instead of duplicating it, the exact block text with placeholders, and a warning about what breaks if somebody edits the block by hand.</format>

💡

Pro tip: Have the script find the block by a marker string rather than by position. Position based scripts break silently the moment somebody adds a paragraph above.

Debug an Apps Script Error

35/35

✨ What it does

Claude explains what the Apps Script error actually means, returns a fixed script, and names the real cause rather than a symptom. You replace the code and rerun on the same failing case.

<context>Here is my script: [paste the code]. Here is the error and the execution log: [paste both]. It fails [every time / only on large docs / only for other users on the team].</context> <task>Explain what the error actually means, fix the script, and tell me which part was the real cause rather than a symptom.</task> <format>The corrected script, a short root cause explanation, and a list of the quota, timeout, or permission limits that could make this same script fail again later.</format>

💡

Pro tip: Copy the whole execution log from the Executions panel, not just the red error line. The entry immediately before the failure usually names the real problem.

Free tool

Prompt Optimizer

Turn a rough idea into a structured, professional AI prompt.

Try it free →

Frequently Asked Questions

No. Claude does not open, edit, or save your documents, and there is no Claude button inside the Docs toolbar. You copy text out of the doc into the chat and paste Claude's output back in. Some paid plans can connect Google Drive so Claude can read a file you point it at, but that connection is read only and never writes anything back. Every prompt on this page is built for that copy in, paste out loop.
Claude returns markdown by default, and Google Docs can convert it. Turn on Tools then Preferences then Automatically detect Markdown, and pasted headings, bold text, and lists convert as you type them. For anything more complex, ask Claude for tab separated rows and paste those, since Docs turns tab separated text into a real table cleanly.
Far more than a typical business document. A 30 or 50 page doc fits in a single message with room to spare, which is why whole document work like reverse outlining and consistency passes are practical. For a very long doc, paste it once at the start of a conversation and run several prompts against it rather than re-pasting each time, so Claude keeps the same view of the text.
Check your own policy first. Anthropic does not train on business tier conversations by default, and Team and Enterprise plans add admin controls and retention settings, but that does not override an internal rule about where client or personal data may go. When in doubt, strip names, salary figures, and customer identifiers before pasting. Most of these prompts work perfectly well on redacted text.
No. Claude writes the script, and you paste it into Extensions then Apps Script inside a Doc or Sheet, then press Run. The first run asks you to authorize the script, which is normal. Start with a dry run flag that only logs what it would change, because Apps Script edits cannot be undone with Ctrl+Z across files.
They do different jobs. Gemini sits in the sidebar and can act on the open document, which is convenient for short tasks. Claude is stronger on whole document reasoning, long transcripts, reconciling two versions, and writing the Apps Script that automates the work permanently. Many teams use the sidebar for quick edits and Claude for the structural passes and the scripts.

Prompts are the starting line. Tutorials are the finish.

A growing library of 300+ hands-on tutorials on ChatGPT, Claude, Midjourney, and 50+ AI tools. New tutorials added every week.

7-day free trial. Cancel anytime.