ChatGPT prompt library

You're helping me write a work email. Here are my rough notes on what I want to say: [paste your bullet points]. Write it as a clear, friendly-professional email to [who it's going to], about [length, e.g. 120 words]. UK English. Keep my meaning exactly; don't add commitments, dates or details I haven't given you. End with a clear next step. Give me a subject line too.

Why this works: It supplies role, audience, tone and a length, and forbids invention, so instead of a generic template you get a real draft built only from what you provided, ready to check and send.

Help me reply to this message: [paste the message]. The situation: [what's actually going on]. I want to sound [tone, e.g. calm, sincere, firm but polite], acknowledge their point, and offer [what you can actually offer]. Draft a reply of about [length]. Don't over-promise or commit to anything I haven't said I can do. Flag anything you're unsure I'd want to include.

Why this works: Naming the situation, the tone you want and the outcome you can actually offer lets it write a real reply rather than a fill-in-the-blanks shell, and the 'don't over-promise' guard keeps you out of trouble.

Here's a draft I've written: [paste it]. First, in three bullet points, tell me what's weak about it: clarity, tone, or length. Then give me a tightened version that fixes those, keeping my voice and meaning. UK English. Note in one line what you changed and why, so I can decide whether I agree.

Why this works: Asking for a critique plus a revision, and to show what changed, teaches you as it edits, so you see the weaknesses, not just a smoother version you can't learn from.

Rewrite this message for a specific reader: [paste the message]. The reader is [who they are and your relationship, e.g. a busy senior colleague, a frustrated customer, a new team member]. Match the right level of formality and detail for them. Give me two versions, one slightly warmer, one more concise, so I can pick. Keep the facts identical.

Why this works: One draft rarely fits every audience; naming the reader and the relationship gets you a version pitched correctly, and asking for two options lets you choose rather than accept.

Summarise the document below for someone who has to make this decision: [state the decision]. Give me: (1) a three-sentence overview, (2) the key points that bear on the decision as short bullets, and (3) anything important the document leaves unanswered. Don't add outside information; summarise only what's here, and say if something's unclear. Document: [paste].

Why this works: A summary aimed at a decision is more useful than a neutral one; naming the decision and asking for what's missing turns a précis into something you can act on.

From this email thread, extract every action or commitment as a table: Action, Owner, Deadline. If an action was mentioned but nobody was clearly assigned, mark the owner as UNASSIGNED. Ignore general discussion; only things someone has to do. If a deadline was implied but not stated, write "not specified" rather than guessing. Thread: [paste].

Why this works: Long email threads bury who-owes-what; asking for owners and flagging the unassigned ones surfaces exactly the items that otherwise slip.

Summarise the material below two ways. First, a two-sentence version for a senior leader who wants only the headline and the one thing that needs their attention. Second, a short bulleted version for the team who'll act on it, with the practical detail. UK English, no invented facts. Material: [paste].

Why this works: The same material often needs a headline for leaders and detail for the doers; asking for both in one pass saves a second round and keeps them consistent.

Here's a data file: [attach or paste]. Before any totals, give me a short data-quality report: blank cells, values in an odd format, categories that look misspelled or duplicated, and any row where the numbers don't add up. Calculate, don't estimate. Then wait; I'll tell you how to handle each issue before you produce any figures.

Why this works: Problems-before-totals catches typos, blanks and mismatches that would otherwise flow silently into your headline numbers; insisting on real calculation stops a confident guessed sum.

Look at this data: [attach or paste]. Tell me: (1) the two or three most important things it shows, in plain English, (2) anything that looks surprising or worth digging into, and (3) what this data can't tell me: the questions it doesn't answer. Base every point on the actual figures, and don't infer causes the data doesn't support.

Why this works: Asking for the story plus the caveats, and what the data can't tell you, produces honest analysis rather than a confident narrative that outruns the evidence.

I'm about to argue the following: [state your position or plan]. Act as a thoughtful sceptic. Give me the three strongest objections someone might raise, the weakest part of my reasoning, and one thing I may have overlooked. Be direct; I want the holes found now, not agreement. Then suggest how I'd address the most serious one.

Why this works: Using ChatGPT as a critic rather than a cheerleader is where it earns its keep on decisions; asking for the strongest counter-argument exposes the weak spot before someone else does.

Help me compare these options: [list the options, e.g. suppliers, tools, approaches]. Put them in a table with these columns: [your criteria, e.g. cost, time, risk, effort]. Fill each cell from what I've told you; where I haven't given you something, write "need to find out" rather than guessing. Then give me a one-paragraph recommendation with your reasoning, so I can challenge it.

Why this works: A structured comparison forces the same criteria across each option instead of a lopsided pros-and-cons ramble, and asking for a recommendation with reasoning makes it decision-ready.

I've got a meeting about [topic] with [who]. Here's the context: [paste notes or background]. The outcome I want is [what a good result looks like]. Give me: a short agenda, the two or three decisions we actually need to reach, and three sharp questions that would move things forward. Keep it to one screen.

Why this works: Turning scattered context into an agenda, the decisions needed and the questions to ask means you walk in ready; naming the outcome you want keeps it focused on results, not just topics.

Turn these rough meeting notes into a clean summary: [paste notes]. Structure it as: a two-line summary at the top, then Decisions (one line each), then Actions as a table (Action, Owner, Due date, marking any with no owner as UNASSIGNED). Don't invent anything that isn't in the notes; if something's ambiguous, flag it rather than smoothing it over.

Why this works: It imposes a consistent, skimmable structure on a rough capture and separates decisions from actions, so the write-up is usable minutes rather than a transcript you still have to mine.

Using the meeting notes below, draft a short follow-up message to the team: what we decided, then each person's actions with deadlines. Warm and brief, UK English. Make it understandable to someone who wasn't there. Flag with "[check]" anything you're inferring rather than certain about, so I can confirm it before sending. Notes: [paste].

Why this works: Going straight from notes to a sendable follow-up captures the meeting while it's fresh; making it stand alone for absentees means one message serves everyone.

I need to [describe the goal, e.g. roll out a new feedback process]. Break it into a sensible sequence of steps, from the very first thing I should do to done. For each step, note who'd need to be involved and roughly how long it might take. Flag any step that depends on another finishing first. Then tell me the single thing to do this week to get moving.

Why this works: A big fuzzy job stalls because you can't see the first move; asking for sequenced steps with a clear starting point turns 'I should sort this out' into something you can begin today.

Here's everything on my plate right now: [paste your list]. Help me prioritise. Sort it into: do first (high impact, quick), schedule (high impact, slower), delegate if possible, and drop or defer (low impact). Be honest about what probably isn't worth doing. If anything looks urgent-but-not-important, say so. Then suggest a realistic focus for today.

Why this works: Asking it to sort by impact and effort, and to name what to drop, forces genuine prioritisation instead of a tidied-up version of the same overwhelming list.

Help me draft a lightweight plan for [project]. The goal is [objective], by [deadline], and the key deliverables are [list them]. Give me: a short timeline broken into phases, who owns each phase [or leave as UNASSIGNED], the main risks with a one-line mitigation each, and any assumptions I'm making that could be wrong. Keep it to a page; this is a working plan, not a formal document.

Why this works: Naming the deliverables, the timeframe and the risks up front produces a plan you can actually share, and the assumptions section keeps it honest about what's still unknown.

I'm weighing up a decision: [describe it, and the options you're considering]. Lay out each option with its main trade-offs: what I gain and what I risk. Add any option I might not have considered. Don't tell me what to do; instead, tell me what additional information would most change the answer, so I know what to find out before deciding.

Why this works: Framing it as options-with-trade-offs rather than 'what should I do' keeps you in charge of the call while using ChatGPT to surface angles you'd miss; asking what would change your mind guards against false confidence.