Prompt Library

ChatGPT Prompts for Dungeon Masters: NPCs, Hooks and Recaps

30 copy-paste prompts

Copy-ready prompts for tabletop game masters: NPCs, plot hooks, encounters, loot, session recaps and homebrew rules, built around your own original campaign.

In short: This page contains 30 copy-paste ready prompts, organized into 6 categories with a description and pro tip for each. The first 5 prompts are free instantly, no signup needed. Drafted with AI writing assistance and checked against our editorial policy.

Louis Corneloup
By Louis Corneloup · Founder, Techpresso
Last updated ·Drafted with AI assistance (how we make prompts)

NPCs and Villains

5 prompts

Tavern Keeper With a Secret

1/30

✨ What it does

Builds one tavern keeper with a visible trait, a hidden secret and a reason to help the party. Check the secret against your campaign notes before you place it in a scene.

Act as a game master's NPC designer for my [GAME SYSTEM NAME] table, set in my original world [SETTING NAME]. Campaign notes: [CAMPAIGN NOTES]. Give the tavern keeper a name, one visible trait, one secret the party can uncover in two scenes, one reason they would help the group and a line they say when nervous. Treat the name, trait, secret and nervous line as drafts I can accept or cut. Do not invent facts about my world beyond [CAMPAIGN NOTES]; tag those gaps [VERIFY]. Output 5 bullets, under 150 words.

💡

Pro tip: Name the town in your notes first. A keeper tied to a real place in your world feels less like filler.

Villain Motive Score

2/30

✨ What it does

Scores a villain draft on motive, leverage and reach, then rewrites the motive using only facts you already set. Read the score as a prompt to revise, not a verdict on the villain.

Read my villain draft below and score it from 1 to 5 on motive, leverage and how the party can reach them: [PASTE VILLAIN DRAFT]. Then rewrite only the motive in 60 words, keeping every fact I already set in [CAMPAIGN NOTES]. Do not invent new history for the villain; flag any gap as [VERIFY] so I can fill it myself. Output a score table with three rows, Motive, Leverage and Reach, then the rewritten motive.

💡

Pro tip: Paste the full draft, including the villain's goal. A short draft gets low scores for reasons that are not really there.

Shopkeeper Interview Build

3/30

✨ What it does

Asks five questions first, then turns your answers into a one-page shopkeeper card with a voice tip. Read the card aloud once before the session to check the voice fits.

Ask me exactly 5 questions about the shopkeeper I want to run in [TOWN NAME]: what they sell, what they fear, how they treat strangers, what they owe and who they trust. Wait for my answers before you write anything. Then build a one-page NPC card with a voice tip for my table. Keep the NPC card original; never use names from a published book or game. Do not invent backstory my answers do not support; if I skip what they owe, write [VERIFY] on that line.

💡

Pro tip: Answer in one or two sentences each. Short answers give the card cleaner facts.

Rival Adventurer Table

4/30

✨ What it does

Lays out four rival adventurers in a table with a goal, a clash with the party and a return hook. Use the Missing fact column as your list of things to decide before the next session.

Build a table of four rival adventurers for my [CAMPAIGN NAME] set in [SETTING NAME]. Columns: Name, Goal, Friction with the party, Hook to bring them back, Missing fact. Setting notes: [SETTING NOTES]. The names, goals, frictions and hooks are new ideas you may propose inside the setting that [SETTING NOTES] describes. Do not invent setting facts beyond it; tag those gaps [VERIFY]. Never use names from a published book or game. In Missing fact, list what I still need to decide about that rival, or write [VERIFY]. Keep the table to four rows and under 200 words.

💡

Pro tip: Give each rival a different goal. Four rivals who want the same thing blur into one character.

NPC Voice Checklist

5/30

✨ What it does

Gives a six item voice checklist for one non-player character, from pace and tics to the question they always ask. Practise the checklist once in your head before the session so the voice is ready when players arrive.

Make a checklist I can use to voice [NPC NAME] at the table. Include 6 items: accent or pace, a verbal tic, how they greet the party, a gesture, a line they never say and one question they always ask. Make up the accent, tic, gesture and lines yourself, then tell me what I must confirm from [NPC NOTES]. Do not invent history beyond those notes, and tag any history you rely on [VERIFY]. Keep every item under 15 words.

💡

Pro tip: Pick two items from the checklist and keep them steady all session.

Plot Hooks and Quests

5 prompts

Three Hook Options

6/30

✨ What it does

Offers three plot hooks with stakes and a first clue, then picks the one that best fits your notes. Test the pick by asking whether your players already care about the stakes it names.

Give me three plot hooks that pull [PARTY LEVEL] characters into [LOCATION] in my original setting. For each hook, write one sentence on the stakes and one on the first clue. Then pick the hook that fits my notes in [CAMPAIGN NOTES] best and give one reason. Make up the hooks and first clues yourself, then tell me what I must confirm from my notes. Do not invent facts about my world beyond [CAMPAIGN NOTES]; tag those gaps [VERIFY]. Output three bullets, then a final line naming the pick.

💡

Pro tip: Run the prompt twice with different locations. Compare the picks before you commit to one.

Five Step Quest Chain

7/30

✨ What it does

Turns one party goal into a five step chain, with an obstacle on every step and a payoff at the end. Walk the chain once and check that the first four rewards lead onward and that step five pays off.

Build a quest chain in five numbered steps for a party that wants [PARTY GOAL]. Each of the first four steps needs a location, an obstacle and the reward that moves them to the next step. Step five needs a location, an obstacle and the payoff. Set it only in [SETTING NAME], a world I made myself, and keep it clear of any published adventure. Places I have: [PLACE LIST]. Use only these places; invent only the obstacles and rewards. If [PLACE LIST] has fewer than five places, reuse a place and say so. If a place from my list is too vague to use, write [VERIFY] on that step rather than adding a new one. Keep each step under 40 words.

💡

Pro tip: Make step four the hardest, so step five pays off the climb.

Rumour Seed Options

8/30

✨ What it does

Offers two options each for who spreads a rumour, the first clue and when players hear it, then marks a pick per field. Check every place or name tagged [VERIFY] against your town notes before you seed the rumour.

Here is a rumour I want to seed in [TOWN NAME]: [ONE LINE RUMOUR]. The truth behind it is: [TRUTH]. Using only [TOWN NOTES], write two options each for Who spreads it, First clue and When players hear it. Mark one option per field as your pick and give a one line reason for each pick. If a place or name is missing from my notes, write [VERIFY] there rather than inventing it. Output the three fields with both options and the pick marked, and stay under 160 words.

💡

Pro tip: Write the truth in one plain sentence first. The rumour is easier to spread when you know what is really going on.

Side Quest Recap Rewrite

9/30

✨ What it does

Rewrites a side quest into a 90 word recap with the reward first and the quest giver named. Check the rewrite against your original, since a short recap can drop a detail that matters later.

Rewrite my side quest below for a player who skipped the last session: [PASTE SIDE QUEST]. Keep it to 90 words, put the reward in the first sentence and name the one NPC who gives the quest. Use only facts from my paste and do not invent names, places or deadlines. If my paste names no reward, write [VERIFY] as the first sentence's reward. If my paste names no quest giver, write [VERIFY] rather than inventing an NPC. Output a short recap paragraph, then one line saying what to do first.

💡

Pro tip: Paste the quest as you wrote it, including the messy parts. The rewrite keeps only what you gave it.

Six Week Story Arc

10/30

✨ What it does

Maps a six week arc into one scene, one choice and one tracked consequence per week. Read the consequence column first, because it shows whether the choices carry forward.

Plan a six week story arc for [PARTY NAME] at [SESSIONS PER WEEK] sessions per week, building toward [CLIMAX]. Give one row per week: the scene, the choice the players face and the consequence I will track. Draft the scenes, choices and consequences as proposals I can keep or drop. Do not invent past events beyond [CAMPAIGN NOTES]; tag references to them [VERIFY]. Output a six row table with columns Week, Scene, Choice and Consequence.

💡

Pro tip: Fill the consequence column before the scenes. Choices that change nothing later tend to feel empty at the table.

Encounters and Combat

5 prompts

Encounter Felt Off

11/30

✨ What it does

Lists three likely causes for an encounter that felt wrong, each with a fix you can try next session. Check any number the fixes depend on against your rulebook before you change the encounter.

My encounter felt [TOO EASY OR TOO DEADLY] at [PARTY LEVEL] with [NUMBER OF PLAYERS] players. Encounter setup: [ENCOUNTER SETUP]. Here is what happened: [SESSION SYMPTOMS]. List the three most likely causes in plain words, then one fix for each. If a fix names a creature, give it a name you make up for this table. Mark every rule threshold or stat value you suggest as [VERIFY] against my own rulebook, and do not invent monster stats. Output a numbered list of three causes with their fixes.

💡

Pro tip: Describe what the players did, not how you felt. Concrete moves give the causes better aim.

Battle Site Description

12/30

✨ What it does

Writes an 80 word battle site with size, light, a hazard, cover and an exit under clear headings. Read it aloud once and cut any sentence you stumble over.

Describe the battle site for [ENCOUNTER NAME] in [SETTING NAME]: size: [GRID DIMENSIONS] (for example 30 by 40 squares), light, one hazard, two cover spots and one exit. Keep it to 80 words I can read aloud. The light, hazard and cover spots are new ideas you may propose inside the area [MAP NOTES] describes. Do not invent rooms or landmarks beyond it. If [MAP NOTES] names no exit, write [VERIFY] on Exit. Output a labelled block with the headings Size, Light, Hazard, Cover and Exit.

💡

Pro tip: Give the grid as dimensions, such as 30 by 40 squares, in the units your map uses.

Foe Swap Options

13/30

✨ What it does

Proposes three original replacement foes with tactics, then ranks them from best fit to weakest. Confirm the top pick fits your party size, then look up its stats in your own rulebook.

Suggest three original creatures to replace the [FOE TO REPLACE] in my fight. Setting notes: [SETTING NOTES]. For each, give a one line tactic and a reason it suits [PARTY SIZE] players. Rank the three from best fit to weakest and say why the top one wins in 25 words. Coin each creature at this table; do not invent a stat value, and if [SETTING NOTES] gives no climate or culture cue, write [VERIFY] on that row rather than guessing. Output a ranked list with Rank, Creature, Tactic and Why it fits.

💡

Pro tip: Ask for foes that match your setting's climate or culture. Generic creatures feel pasted on.

Ambush Scene Checklist

14/30

✨ What it does

Gives a seven item ambush checklist, from the trigger and sound cue to the retreat rule. Tick each item off in order during prep so nothing is missing when the ambush starts.

Write a checklist for running an ambush scene for [PARTY NAME] at [LOCATION]. Include 7 items: the trigger, what the foes do in round one, the escape route, the sound cue, the first clue, the retreat rule and the reward for listening. Treat the foes' round-one actions as my table's to accept or cut, and invent the sound cue and first clue yourself. Do not invent extra foes or terrain beyond [SCENE NOTES]. If [SCENE NOTES] gives no escape route, write [VERIFY] there. Output a numbered checklist, no more than 12 words per item.

💡

Pro tip: Write the sound cue on a card so you can read it the same way every time.

Stealth Versus Open Fight

15/30

✨ What it does

Compares a stealth encounter with an open fight on running time, fairness and story, then gives a short verdict. Use the verdict as a starting point and check it against how your table actually plays.

Compare a stealth encounter and a head on fight for [ENCOUNTER GOAL] with my players, who are [PLAYER TYPES]. Cover how long each takes to run, the chance players feel cheated and the story each one leaves behind. Give your judgment on run time, fairness and story as suggestions. Session notes (player habits at the table, such as scouting or rushing in): [SESSION NOTES]. Use [SESSION NOTES] only for facts about my players and do not invent player preferences; tag each guess [VERIFY]. Output a three row table with columns Factor, Stealth and Open fight, then a two sentence verdict.

💡

Pro tip: Describe your players by what they do at the table, such as scouting or rushing in. That is more useful than a label.

Like these prompts? There are full tutorials behind them.

Learn the workflows, not just the prompts. Easy-to-follow tutorials inside AI Academy, and the library keeps growing.

Try AI Academy Free

Loot and Rewards

5 prompts

Treasure Table Draft

16/30

✨ What it does

Drafts a six row loot table with a story hook, a rival claimant and a downside for each item. Check the Value to check column against your rulebook, since the values are marked for checking.

Build a loot table with six rows for a [LEVEL BAND] party in [SETTING NAME] using [GAME SYSTEM NAME]. Columns: Item, Story hook, Who wants it, Downside and Value to check. Keep items original and tie each one to [CAMPAIGN NOTES]. Item names, hooks, claimants and downsides are new ideas you may propose; only rules and amounts must stay [VERIFY]. Do not invent magic rules or amounts; write [VERIFY] in that column. Output the table, then one sentence on pacing.

💡

Pro tip: Give every item a downside so players have a real choice.

Cursed Item Card

17/30

✨ What it does

Offers three story costs and a rival claimant for a cursed item, then picks one and writes a card under 120 words. Read the cost line with your players in mind, since a cost they never feel does little work.

Build a cursed item card for [ITEM NAME], whose power in story terms is [WHAT IT DOES]. Offer three story costs the wielder pays and one rival claimant drawn from [CAMPAIGN NOTES], then pick the cost that creates the most table tension. Keep the power story based, with no mechanics or dice values. Use original item names only, never names from a published book or game. If my notes name no rival, write [VERIFY] rather than inventing a person. Output the completed card in under 120 words.

💡

Pro tip: Describe the power in what it does for a character, not in numbers. Numbers can go in your rulebook later.

Reward List Critique

18/30

✨ What it does

Gives a one line verdict for every reward and rewrites the three weakest with a story reason. Compare the rewrites with the party facts before you hand them out.

Read my reward list below for [GAME SYSTEM NAME] and say which items are too strong, too weak or too similar: [PASTE REWARD LIST]. Give a one line verdict for each item, then rewrite the three weakest with a story reason. Give the rewrites names that exist only at this table. Use the party facts in [PARTY NOTES] only and do not invent stats. Flag any balance number as [VERIFY] against my rulebook. Output a three column table: Item, Verdict and Rewrite.

💡

Pro tip: List the items with their current effect. Verdicts on vague names are guesses.

Loot Handout Checklist

19/30

✨ What it does

Runs six checks on a loot handout plan and marks each as yes, no or verify. Fix every no before the session and settle each verify with a note in your prep file.

Check my loot handout plan before the session: [PASTE LOOT PLAN]. Tick six checks: each item has a story reason, nobody is left out, the cost is clear, the item can be sold or traded, the clue is visible and the rule source is named for [GAME SYSTEM NAME]. Mark any check you cannot confirm from my plan as [VERIFY]. Do not invent items or rules. Output a checklist with a yes, no or VERIFY mark on each line.

💡

Pro tip: List each player's character in the plan so the 'nobody left out' check has something to test.

Three Reward Options

20/30

✨ What it does

Offers three rewards for a rescue, each with a reason it fits your notes, then names the strongest. Check that the pick actually advances a thread from your campaign notes.

Give three reward options for players who just saved [NPC NAME] from [THREAT]. Each option gets a one sentence reward and a reason it fits [CAMPAIGN NOTES]. Pick the one that moves the story forward most and say why in two lines. Name each reward as something that exists only in this campaign. Do not invent names or places I did not give. If [CAMPAIGN NOTES] name no open thread, write [VERIFY] on the pick reason. Output three numbered options, then a final line naming the pick.

💡

Pro tip: Offer one reward that helps the party and one that creates a problem. Both tend to get good discussion.

Session Recaps and Notes

5 prompts

Player Recap Draft

21/30

✨ What it does

Turns rough session notes into a 150 word past tense recap that ends with an open question. Read it against your notes before you post it, since a recap is easy to tidy into something that never happened.

Turn my messy session notes into a 150 word recap for players: [PASTE SESSION NOTES]. Write it in the past tense, name the party as [PARTY NAME] and end with one open question. Use only what I wrote. If I did not say who won the fight, write [VERIFY] rather than inventing the outcome. Output a title line, then the recap.

💡

Pro tip: Paste the notes raw, including the jokes, so the recap keeps the table's own tone.

Cold Open Interview

22/30

✨ What it does

Asks five questions about last session, waits for answers, then writes a 100 word read-aloud cold open for next session. Read the cold open aloud once and cut any beat you did not confirm in your answers.

Ask me exactly 5 questions about [SESSION DATE] before you write anything: the turning point, the NPC who surprised us, what the party lost, the next goal and any clue I missed. Wait for my answers. Then write a 100 word read-aloud cold open for next session in present tense. Do not invent answers I did not give; if I skip a question, write [VERIFY] on that beat.

💡

Pro tip: Answer the turning point question first. It sets the opening image of the cold open.

Session Log Table

23/30

✨ What it does

Logs a session as a table of scenes, actions, changes and open threads, with at most eight rows. Use the open thread column to plan the next session's first scene.

Turn my notes into a session log table with columns Scene, Who acted, What changed, Open thread and Missing detail. Use [SESSION NOTES] only and list each scene in order. If there are more than 8 scenes, merge the smallest and say which. Do not invent events. Mark any blank as [VERIFY] for me to fill. Output a table with at most 8 rows.

💡

Pro tip: Keep one log per session in the same table format. Comparing logs shows which threads you keep dropping.

Recap Clarity Score

24/30

✨ What it does

Scores a recap on clarity, pace and hooks, then rewrites two confusing lines as before and after pairs. Check each rewrite against your notes before you swap it in.

Score my session recap below from 1 to 5 on clarity, pace and the hooks it leaves for next time: [PASTE RECAP]. Then list the two lines that would confuse a player who missed the game and rewrite them. Do not invent events or names that are not in my draft. Flag any gap as [VERIFY]. Output a score line, then two before and after pairs.

💡

Pro tip: Read the recap aloud before you paste it. The lines that trip your voice are often the ones that confuse readers.

Next Session Prep Checklist

25/30

✨ What it does

Gives an eight item prep checklist covering recap, threads, voices, maps, handouts and the cliffhanger. Work down it the afternoon before the session, leaving the cliffhanger item last.

Make a prep checklist for my next session of [SESSION LENGTH] on [SESSION DATE]. Include 8 items covering the recap, open threads, NPC voices, maps, handouts, a safety check in, timing that fits [SESSION LENGTH] and the cliffhanger. Build it from [PREP NOTES] only and do not invent prep items the notes contradict. If no map is named in [PREP NOTES], write [VERIFY] on the maps item. Output 8 checkboxes, each under 12 words.

💡

Pro tip: Set the timing item to a real clock time. A vague timing item tends to get skipped on busy nights.

Go from copy-pasting to actually mastering AI.

AI Academy: hands-on tutorials on ChatGPT, Claude, Midjourney, and many other tools. New tutorials are added regularly.

Start Your Free Trial

Homebrew Rules and Session Zero

5 prompts

Session Zero Questions

26/30

✨ What it does

Creates six numbered session zero questions on tone, lines and veils, pace, death rules and table etiquette. Send the questions a few days ahead so players can think before they answer.

Write a session zero questionnaire with 6 numbered questions for a new table of [NUMBER OF PLAYERS] players at [EXPERIENCE LEVEL]. Cover tone, lines and veils, combat pace, death rules, what to do about a missed session and table etiquette including phones. Keep the wording friendly and plain. Invent no house rules; tag any rule question that needs my rulebook as [VERIFY]. Output a numbered list only.

💡

Pro tip: Ask about lines and veils before tone, so the boundaries are set before the style talk.

Homebrew Rule Card

27/30

✨ What it does

Drafts Limit and Exception for a homebrew rule, then adds a Conflicts line against the rule it replaces. Test the card in one short scene before you make it table law.

Draft a homebrew rule card for [RULE NAME]. Trigger: [WHEN IT APPLIES]. Effect in words: [WHAT CHANGES]. Propose the Limit in words, such as once per scene or session, and tag any number you are unsure of [VERIFY]. Draft an Exception (when it does not apply) that fits the trigger and effect. Name the rule in [REPLACED RULE NAME] and list at most 2 lines of conflicts. Keep the card under 120 words. Do not invent numbers or dice values. Output the card, then the Conflicts line.

💡

Pro tip: Give a tight trigger. Vague triggers make the Limit and Exception hard to draft.

Rule Argument Fix

28/30

✨ What it does

Names three likely causes for a recurring rules argument and proposes one house ruling to test. Try the ruling for a single session, then decide whether to keep it.

My table keeps arguing about [RULE TOPIC] in [GAME SYSTEM NAME] during play. Here is what happens: [DESCRIBE THE ARGUMENT]. Give three likely causes in plain words, then one house ruling to try for one session. Do not invent official wording or quote any rulebook. Name the rule topic I should look up in my rulebook, tagged [VERIFY]. Do not give page or rule numbers. Output a numbered list of three causes and a single ruling, under 150 words.

💡

Pro tip: Describe the argument as a sequence of moves. The point where players disagree is usually the gap in the rule.

Two Versions Compared

29/30

✨ What it does

Compares two versions of a homebrew rule on speed, fairness and clarity, then makes a one line pick. Playtest the picked version for one session before you rely on the verdict.

Compare two versions of [RULE NAME] for my table. Version A is [VERSION A IN WORDS] and version B is [VERSION B IN WORDS]. Compare speed of play, fairness to [PLAYER TYPES] and how easy the rule is to explain. Use my notes only and do not invent playtest results. Mark any claim I have not tested as [VERIFY]. Output a three row table with columns Factor, Version A and Version B, then a one line pick.

💡

Pro tip: Write each version in one paragraph. Short versions show the difference more clearly than long ones.

Plain English Rule Rewrite

30/30

✨ What it does

Rewrites a homebrew rule for a new player in under 90 words, with one example line. Read the rewrite to someone new to the game and ask them to repeat the rule back.

Rewrite my homebrew rule below for a new player who has never played [GAME SYSTEM NAME]: [PASTE RULE TEXT]. Use short sentences, no more than two game terms, and a one line example. Keep it under 90 words. Do not invent rules my text does not contain. Tag any term you are unsure of as [VERIFY]. Output the rewrite, then the example line.

💡

Pro tip: Read the rule aloud to a non-player first. Terms you stopped noticing are the ones that need rewriting.

Free tool

Prompt Optimizer

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

Try it free →

Frequently Asked Questions

It can suggest ideas, but it does not know your world. Paste your campaign notes; the prompts ask ChatGPT to tag missing facts [VERIFY]; read the output for gaps it did not tag.
No. Several prompts ask for original names, creatures and items and tell ChatGPT to avoid published ones. Check any output yourself before you use it, and check a published setting's terms if you build on one.
Several prompts, such as Encounter Felt Off and Reward List Critique, tell ChatGPT to mark stat values [VERIFY]. It can still miss one, so check them against your rulebook.
Use first names or made up handles for players and leave out personal details such as addresses or work information. Anything you paste is shared with the model, so keep private matters out of the notes.
Replace the [GAME SYSTEM NAME] bracket where it appears and check every mechanic against your own rulebook. The story parts, such as NPC voices and recaps, carry over to other systems with little change.

Prompts are the starting line. Tutorials are the finish.

A growing library of hands-on tutorials on ChatGPT, Claude, Midjourney, and many other AI tools. New tutorials are added regularly.

7-day free trial. Cancel anytime.