Prompt Library

Save, reuse, and manage your favorite prompts across chats.

Written By Mark Ku

Last updated 2 days ago

How to open the Prompt Library

You can open the Prompt Library from the chat interface in two ways:

From the prompt bar:

  1. In the chat, look at the area above the prompt input.

  2. Click the book icon (Prompt Library) next to the scope pill (e.g. “Chatting with…”).

  3. The Powering This Chat side panel opens with the Prompts tab selected.

From the side panel:

  1. Open the Powering This Chat panel (from the chat header or the side-panel toggle).

  2. Click the Prompts tab.

  3. You’ll see your saved prompts and a Create New Prompt button.

💡 Tip: If you have text in the prompt bar and click the book icon (or open the panel and go to Prompts), you can start Create New Prompt and your current prompt text will be pre-filled. Great for turning a one-off prompt into a reusable template.

Overview

The Prompt Library is a place to store prompts you use often. Instead of retyping or copying from previous chats, you save a prompt once—with an optional name and description—and reuse it in any chat. You can run the same prompt across different collections, files, and projects without digging through history.

Benefits:

  • Consistency — Run the same analysis or question format every time.

  • Speed — One click to load a prompt into the prompt bar.

  • Flexibility — Use template variables and @mentions so one prompt can adapt to different topics or files.

  • Organization — Search, star, and sort your prompts so favorites stay easy to find.

Creating a new prompt

You can create a stored prompt from scratch or from text you’ve already typed.

To create a new prompt:

  1. Open the Prompt Library (see How to open the Prompt Library).

  2. Click Create New Prompt.

  3. In the modal:

    • Name — Enter a short, recognizable name (e.g. “Weekly summary” or “Competitive analysis”). This is required.

    • Description — Optionally add a sentence so you can tell prompts apart when searching.

    • Prompt — Enter the prompt text. You can use @ to mention files, collections, labels, concepts, or project knowledge, and you can use template variables (see Template variables and @mentions).

  4. Click Create New Prompt again (the button at the bottom) to save.

The new prompt appears in your library. You can search for it by name or description and click it to load it into the prompt bar.

Saving a prompt from chat

When you’ve written a prompt in the chat (or sent one and want to reuse it), you can save it to the Prompt Library without retyping.

To save a prompt from a message in a chat thread:

  1. Find the message that contains the prompt you want to reuse.

  2. Use the Save to Prompt Library option on that message

  3. Fill in Name and optional Description in the modal, then save.

After saving, you can open the Prompt Library and apply this prompt to the prompt bar in any chat.

💡Tip: Saving from the prompt bar preserves @mentions and template variables. When you apply the stored prompt later, those placeholders are restored so you can adjust them and send.

Using prompts from your library

To use a stored prompt in the current chat:

  1. Open the Prompt Library

  2. Optionally use the search box to filter by name or description.

  3. Click the prompt you want to use. The prompt is inserted into the prompt bar

  4. The side panel closes and focus moves to the prompt bar. Adjust scope, add or change @mentions or variables if needed, then send.

Clicking a prompt applies it; it does not send it automatically. You can edit the text before sending. Starred prompts appear at the top of the list so you can get to them quickly.

Managing prompts: star, edit, delete

  • Star / Unstar — On each prompt row, use the star icon to mark favorites. Starred prompts are listed first and are easy to spot. Toggle again to unstar.

  • Edit — Use the pencil icon on the row to open the edit modal. You can change Name, Description, and Prompt text. Save to update.

  • Delete — Use the trash icon to delete a prompt. You’ll be asked to confirm. This action cannot be undone.

All of these actions are available from the prompt row (hover to see star, edit, and delete). The list is sorted with starred prompts first, then alphabetically by name.

🚀 Pro-Tip: Make your prompts dynamic

You can make your prompts powerful and reusable by combining @mentions and template variables (#variables). This lets you focus prompts on the right context while adapting them for multiple scenarios, saving time and keeping your responses consistent.

Why it matters:

  • One library prompt can be reused across different datasets.

  • Perfect for recurring analyses, weekly reports, or cross-project summaries.

  • Keeps your workflow fast, consistent, and adaptable.

Using #variables

You add template variables while writing or editing a prompt (in the prompt bar or in the Create / Edit prompt modals).

  1. In the prompt text area, type #.

  2. A popup appears with Add template variable. Click it.

  3. In the modal that opens:

    • Choose a type — Text Input, Paragraph Input, File Picker, Record Audio, or Dropdown (see Template variable types below).

    • Name — Enter a short label for the variable (e.g. “Topic”, “Report”, “Detail”). This label appears as the placeholder in the prompt so you know what to fill in.

    • For Dropdown only: add the options users can choose from (e.g. “Brief”, “Detailed”, “With Examples”).

    • Optionally mark the variable as optional and set a fallback value used when it’s left empty.

  4. Click Create (or the equivalent confirm button). The variable is inserted into the prompt.

When you apply a stored prompt that contains template variables, those placeholders appear in the prompt bar as fillable fields. You fill them in (or leave optional ones empty), then send. This lets one prompt work for many situations—e.g. “Summarize” plus a file variable for different reports.

Template variable types

Type

What it does

Text Input

A single-line text field for short, free-form input (names, topics, keywords).

Paragraph Input

A multi-line text area for longer input (descriptions, instructions, context).

File Picker

Lets the user select a file from their project assets (documents, spreadsheets, etc.).

Record Audio

Lets the user record a short audio clip that is uploaded and used as context.

Dropdown

A fixed set of options the user chooses from (e.g. output format, detail level).

  • Text and Paragraph are best for open-ended input (a topic, a question, a description).

  • File Picker is best when the prompt should run on a specific file the user chooses each time (e.g. a “Summarize” prompt with a file variable).

  • Dropdown is best when you want a constrained choice (e.g. “Brief answer” vs “Detailed explanation”).

  • Record Audio is for adding voice input as context (e.g. instructions or notes).

You can mix types in one prompt—e.g. a file variable for the report, a dropdown for detail level, and a text variable for audience.

Learn more about variables: Creating prompts with variables

Using @mentions

@mentions let you focus your prompt on specific files, Collections, Labels, Concepts, or Project knowledge, instead of searching across all your assets. Stored prompts save these references, and when you apply the prompt later, the same structure is restored, so you can adjust scope as needed before sending.

Steps:

  1. Type @ in the prompt editor (or Create/Edit modal).

  2. Select the items you want from the dropdown (you can mention multiple items).

  3. Each selected item appears as a chip in the prompt bar for verification.

Quick tips:

  • Mention multiple items for side-by-side comparisons or analysis.

  • Give assets and concepts clear names for easy lookup.

  • Use @mentions in combination with template variables (#variables) to create reusable, adaptable prompts.

Learn more about mentions: Using @mentions

🚀 Pro-Tip: Use @ to put specific Views, Labels, Concepts, or files in scope, and use template variables (e.g., #{text:Topic}, #{file:Document}) for the parts that change each time. That way, one library prompt can be reused across different datasets and contexts.