Balane TextDeck

Free AI prompt library: 15 starter templates

Most "free" prompt libraries are lead magnets. This one is the real thing: 15 templates you can use today without signing up for anything.

Updated October 20, 20267 min read15 prompts to copy

Searching for a free AI prompt library usually lands you on a signup wall, a Notion page selling a $29 upgrade, or a ChatGPT marketplace that needs a credit card. This post is neither. Below are 15 genuinely free, ready-to-use, parameterized prompt templates covering the most common daily AI tasks. Copy, paste, use.

What counts as "free" here

Free as in: no signup, no email capture, no premium tier, no login. You can copy these templates, paste them into any AI service, and use them forever. They work in any tool that stores text.

If you want them a keystroke away, TextDeck is free and does that. The app ships 135 curated templates in 14 packs, including a starter pack with professional email, constructive feedback, meeting notes, SWOT and a quiz generator, so these fifteen are a starting point.

What makes a prompt library worth keeping

A useful prompt library has three properties. It is curated: you kept the prompts that produced something you used and dropped the rest. It is parameterized: the parts that change are marked as {{variables}}, so you fill in a topic instead of rewriting a paragraph. And it is retrievable: any template is in front of you within three seconds.

Many free libraries online miss the first point by design. A PDF of 500 prompts is easy to give away and hard to use, because nobody scans 500 entries for the one that fits. The fifteen below cover the tasks that come up most often in daily AI work, and each one already carries its variables.

1. Summarize anything

The summary a model gives you unprompted is a shorter version of everything. Naming a sentence count and adding "skip fluff" turns it into a decision about what matters, which is what you actually wanted.

Prompt

Summarize the following in {{length}} sentences. Preserve key facts. Skip fluff. {{content}}

Try it: fill the variables

2. Explain to an expert in an adjacent field

"Explain like I am five" insults the reader and throws away precision. Naming a field the reader does know gives the model a real vocabulary to borrow from instead of a register to flatten into.

Prompt

Explain {{concept}} to someone with expertise in {{adjacent_field}} but no background in {{target_field}}. Use an analogy they would recognize. Max 200 words.

Try it: fill the variables

3. Turn messy notes into structured output

The "do not invent" clause is doing most of the work here. Without it a model will helpfully fill an empty field with something plausible, and you will not notice until it matters.

Prompt

Convert these messy notes into {{format}} with these fields: {{fields}}. If a field is missing, mark as null. Do not invent. {{notes}}

Try it: fill the variables

4. Draft an email

The one specific detail is what separates a usable draft from something that reads like a mail merge. Everything else in an email a model can guess; that it cannot.

Prompt

Draft a {{tone}} email from {{sender_role}} to {{recipient_role}}. Purpose: {{purpose}}. Include one specific detail that shows context. Max {{length}} words.

Try it: fill the variables

5. Critical review

Models default to encouragement. "Be direct. Do not soften." is the shortest instruction that reliably gets past it, and you can always ask for the case in favour separately.

Prompt

Review the following {{content_type}} critically. Identify weak claims, unsupported assertions, logic gaps. Be direct. Do not soften. {{content}}

Try it: fill the variables

6. Brainstorm with constraints

Unconstrained brainstorming returns the obvious five. Constraints, a forced ranking, and one deliberately unconventional entry give you something you would not have arrived at alone.

Prompt

Brainstorm {{count}} approaches to {{problem}}. Constraints: {{constraints}}. Rank from most to least feasible. Include one that is unconventional.

Try it: fill the variables

7. Compare options

The markdown table forces the model to fill every cell, which is what exposes the dimensions it does not actually know. The flag instruction turns those into an honest "depends" instead of a confident guess.

Prompt

Compare {{options}} across these dimensions: {{dimensions}}. Output a markdown table. Flag any dimension where the answer depends on context.

Try it: fill the variables

8. Meeting prep

The "two things I should not bring up" line is the part people leave out and the part that pays. It makes the model reason about the other side of the table, not just your side of it.

Prompt

I have a meeting with {{person_or_company}} about {{topic}}. Goal: {{goal}}. Give me: 5 questions to ask, 3 points to make, 2 things I should not bring up.

Try it: fill the variables

9. Learning plan

Hours per week is the variable that changes the whole plan. Leave it out and you get a curriculum built for someone with unlimited time, the kind you abandon in week two.

Prompt

Build a {{duration}} learning plan for {{topic}}. I have {{hours_per_week}} hours per week. Current level: {{current_level}}. Output: weekly milestones, resources, one checkpoint per week.

Try it: fill the variables

10. Decision framework

Ending with a forced recommendation matters. A model asked to "walk through" a decision will happily stop at "it depends" unless you close that exit explicitly.

Prompt

I am deciding between {{options}} for {{context}}. Criteria I care about: {{criteria}}. Walk me through a structured decision. End with a recommendation.

Try it: fill the variables

11. Persona-based rewrite

"Do not add information" is the guardrail. Tone transfer without it drifts into invention, and you end up fact-checking a rewrite of something you already knew was true.

Prompt

Rewrite the following in the voice of {{reference_style}}. Preserve meaning. Do not add information. {{original}}

Try it: fill the variables

12. Extract structured data

Same shape as the notes template, tighter output. Asking for JSON by name gets you something you can parse; asking for "a structured summary" gets you prose with bullet points in it.

Prompt

Extract the following fields as JSON from the text: {{fields}}. Use null for missing fields. Do not invent. {{text}}

Try it: fill the variables

13. Step-by-step plan

Numbering the steps forces the reasoning to happen in order. Without the numbering a model jumps to a recommendation and back-fills the justification, which reads the same and is worth much less.

Prompt

Help me plan {{goal}}. Step 1: constraints. Step 2: assumptions. Step 3: three approaches with tradeoffs. Step 4: recommendation with reasoning. Context: {{context}}.

Try it: fill the variables

14. Uncomfortable questions

The most useful prompt on this list. You already have the case for your decision; what you are missing is the ten questions you have been quietly avoiding.

Prompt

I am considering {{decision}}. Give me the 10 most uncomfortable questions someone skeptical of this decision would ask. No softballs.

Try it: fill the variables

15. Code review

"Hidden state and non-obvious control flow" is the specific ask. Request a general review and you get style notes; request this and you get the bugs.

Prompt

Review the following code for {{concern}}. Focus on correctness, edge cases, readability. Flag any hidden state or non-obvious control flow. {{code}}

Try it: fill the variables

Arrange them in three layers

Layer 1 is your frequent five: the templates you run several times a day, usually a summary, a rewrite, an email, a review and one specific to your job. They belong one hotkey away, with the values you used last already filled in.

Layer 2 is the working set: ten to twenty templates for weekly tasks, sorted into four to six folders by purpose, such as Communication, Analysis, Writing, Development and Research. Most of templates 6 to 13 above end up here.

Layer 3 is the archive: prompts you might need once a quarter. Keep them searchable and out of the way, so they never crowd the first two layers.

Keep the library independent of any one AI service

Every template on this page works in ChatGPT, Claude, Gemini, Copilot or a local model, because none of them depends on a feature of one service. Keep your library the same way. Prompts saved inside one chat account stay behind when you move a task to another model, and a new model release is a bad day to rebuild twenty templates.

A neutral place can be as simple as a text file. A dedicated prompt manager adds what a file cannot: fill-in fields for the variables, a hotkey and search.

Do you need a paid prompt library?

For personal use, rarely. Most paid prompt libraries sell a long list of prompts with a thin interface on top. The prompts are the cheap part: you can collect fifteen good ones in an afternoon, and the ones that pay off over months are those you adapted to your own work and your own variables.

TextDeck is free with every feature included: no account, no subscription, no seat limits. It ships 135 curated templates in 14 packs, runs on Mac, iPhone, iPad, Android and Windows, and keeps your prompts on your device unless you switch on sync.

How to make this a library

Copy these fifteen into any prompt manager. In TextDeck the {{double-brace}} variables become fill-in fields, and every template is one shortcut away: ⌃⇧P on the Mac, Ctrl+Alt+P on Windows, Shortcuts, Spotlight or the Action Button on iPhone, and the text-selection menu on Android.

Then add your own. Every prompt you write twice belongs next to these, with its moving parts marked. The guide to saving and reusing ChatGPT prompts walks through that in four steps, from a chat thread to a named template you can find again.

Next guide

How to improve an AI prompt

A weak prompt and a strong prompt are usually only two sentences apart. Here are the five edits that close the gap, each one shown as a rewrite you can drag through.

Three keys. Every template.