Prompts

📜 Prompts

The Prompts tab in Settings holds the text the app sends to the AI for each request. A prompt preset is one full set of that text, with its per-prompt options. The app sends the prompts of the active preset.

The Prompts tab shows in Advanced mode only. To open it, open Settings, select Advanced in the switch next to the title, then open the Prompts tab.

A prompt sends only what it holds. Text you delete is gone from the request, and a chip you delete sends nothing. See The Chip Editor.

How to Make a Prompt Preset

The built-in presets are read-only. To change a prompt, make your own preset first.

1. Open the Prompts tab.
2. Open the Preset list and select Add New Preset….
3. Type a name in the New Preset dialog.
4. Select Save.

The new preset is a copy of the preset that was active, and it is now the active preset. To copy a preset in one step, select the Duplicate button next to the Preset list, or Duplicate & Edit on a read-only notice. The copy is named "preset name (copy)".

How to Edit a Prompt

1. Select your own preset in the Preset list.
2. In the list of prompts, select the prompt, such as Choices.
3. Under the prompt, select System Prompt, or User Message when the prompt has one.
4. Type in the editor. To add a value from the game, select a chip in the bar above the editor. See The Chip Editor.
5. Open the Preview tab to read the text with sample values in place of the chips.

The app saves each change at once. Under the editor, select Compare to see your changes against the shipped text. To go back to the shipped text, select Reset, then confirm. In the Messages view, each message has its own Reset and Compare beside its name.

How to Route a Prompt to Another Endpoint

You can send one prompt to a different text endpoint, such as a small fast model for Choices. First add the endpoint as a preset on the Endpoints tab. See Text.

1. Select your own preset in the Preset list.
2. Select the prompt, then the Options row under it.
3. Open the Endpoint list and select the endpoint preset.
4. Read the line under the list. Reachable means the endpoint answers. Select Recheck to try again.

To send the prompt to the active endpoint again, select Use Active Endpoint. A prompt route stays on this device. It is never in an exported or published preset.

How to Share a Prompt Preset

1. Select the preset in the Preset list.
2. Select the Export button.
3. Select Copy code to copy a share code, or Export .json to save a file.

To add a preset that someone shared:

1. Select the Import button next to the Preset list.
2. Paste the share code, or select Choose file… and pick the .json file.
3. Read the warnings and the preset's Overview, then set its Name.
4. Select Import. When you selected the checkbox to overwrite a preset with the same name, the button reads Overwrite.

See Sharing a Preset for what a shared preset holds.

How to Publish a Prompt Preset

You must log in to Community Creations to publish. See Login and Register.

1. Select your own preset in the Preset list.
2. Select Overview at the top of the list of prompts.
3. Add at least one model to Models.
4. Fill in Author, Description and Tags.
5. Select Publish, then follow the publish dialog.

Without a model in Models, Publish opens Add a Model. Select Open Overview to go to the field.

How to Use a Preset for One World

1. On the main menu, select the world.
2. In the world dialog, open the Prompts list.
3. Select a preset. Use global preset goes back to the preset that Settings has active.

That world now runs on the preset you picked. Your other worlds do not change. A library Group can also carry a preset, and the list then shows Use group preset. The world's own pick comes first.

While you play a world with its own preset, the Prompts tab edits that world's pick. A note under the Preset list tells you so.

---

Prompt Presets

The Preset list holds four built-in presets, then your own presets.

Preset	What it is
Default	The shipped prompts, with Markdown headings
Simple	The same prompts, with plain labels in place of headings, and chips set to plain text
XML	The same prompts, with each section in an XML tag, and chips set to XML
Experimental	A short narration prompt that sends entity summaries and lets the AI look up full entries. See The Experimental Preset.

A built-in preset is read-only. Its editors show a lock notice with Duplicate & Edit. You can still read every prompt, and you can still switch Tools on and off for it in the Tools tab.

A preset holds:

• the text of every prompt and message
• each prompt's Options, except Endpoint, which only your own presets carry
• which Tools are on. See Tools.
• the Overview, on your own presets

The icon buttons next to the Preset list manage your presets. Point at an icon to see its name. On a narrow screen they are in the Preset Actions menu.

Button	What it does
Duplicate	Makes an editable copy named "preset name (copy)" and selects it. Built-in presets have it too.
Rename	Changes your preset's name
Import	Adds a preset from a share code or a .json file. See How to Share a Prompt Preset.
Export	Opens the share code and the .json download. Built-in presets have it too.
Publish	Lists your preset in Community Creations. Shows while you are logged in.
Reset	Sets every prompt in your preset back to its shipped text
Delete	Removes your preset

Reset and Delete ask you to confirm. Neither can be undone.

The Experimental Preset

Experimental tests a different way to tell the AI about the entities. Its narration prompt sends a one-line summary of each entity, not the full description. The get_entity Tool is on, so the AI can fetch the full entry of an entity it writes about.

On an endpoint with no Tool support, the AI gets the summaries only. The narration then knows less about each entity than with Default. See Endpoints Without Tool Support.

Sharing a Preset

A share code and a .json file hold the same preset:

• the prompt text and the section style
• the tuning: samplers, reasoning, Max Output, Include Attachments and Verbatim Turns
• the Overview
• which Tools are on, and a copy of each of your own Tools that is on

A shared preset never holds a prompt's Endpoint. Endpoint presets exist only on your device.

The Import Preset dialog shows warnings before you import. A preset with a Script Tool warns you, because a script runs code when the AI calls it. Clear the tuning checkbox to import the prompt text only. When you already have a preset with that name, a checkbox offers to overwrite it. Without it, the import adds a separate copy.

The Overview

Overview is the first row in the list of prompts. It shows on your own presets only. It describes the preset for other players when you publish it.

Field	What it holds
Author	Who wrote the preset
Description	What the preset changes and how to use it
Tags	The preset's style and purpose
Models	The models the preset works well with. Publishing needs at least one, because players filter prompts by model.

The Prompts

The list on the left groups the prompts by the job they do. A prompt shows only while its feature is on. When you turn a feature off, its prompt is not in the list, and the app sends nothing for it.

Prompt	Group	What it does in a turn	Shows while
Narration	Story	Writes the story text you read each turn	Always
Planning	Story	Plans the turn before narration writes it: who is present and what happens next	Thinking is Planning
Director	Story	Describes the scene first: who is here and what each entity is doing	Thinking is Staged
Character	Story	Runs once for each staged entity, after the director. States what that entity wants this turn, in the first person.	Thinking is Staged
Discover Entity	Story	Writes a lasting note for each new entity the story names. Runs after narration names a new entity.	Describe New Characters is on
Storyboard	Story	Merges the intentions of every staged entity into one plan, before narration	Thinking is Staged
Choices	Story	Writes your choices for the next turn, after narration	Choices is on
Stat Updates	Trackers	Reads the turn after narration and records which stats changed	Stat Updates is on
Location Change	Trackers	Decides after narration if your action takes you to a new location	Location Change is on
Clock	Trackers	Measures how much story time the turn took. Runs after narration.	Measured Clock is on
Opening	Trackers	Reads the opening scene once, at the start, to set the time of day	Measured Clock is on
Summaries	Memory	Condenses an older turn into one line that the AI can still read later. Runs when a turn is too old to go word for word.	Memory Summaries is on
Milestone Select	Memory	Decides between turns which summaries stay in long-term memory	Memory Summaries is on
Diary	Memory	Writes a private, first-person note from each present entity. Runs when a turn is too old to go word for word.	Character Diaries is on and Thinking is Staged
Scene Tags	Images	Tags the action of the scene for the scene image, after narration	Enable Image Generation is on

The settings in the last column are in Settings → Output, and Enable Image Generation is in Settings → Endpoints → Image. See Settings. A line above each editor repeats what the prompt does.

The Surfaces of a Prompt

Select a prompt to open its Anatomy. The rows under the prompt open its other surfaces. On a narrow screen, one list at the top holds the prompts and the surfaces.

Surface	What it holds	Which prompts have it
Anatomy	The request this prompt sends, built from your current settings	Every prompt
System Prompt	The instructions the AI reads first	Every prompt
User Message	The message that asks for the reply, such as your action	Narration with Thinking set to Native, Choices, Stat Updates, Location Change, Clock, Opening, Summaries, Milestone Select, Director, Discover Entity, Scene Tags
Messages	Extra lines in the narration history	Narration, while one of its messages is live
Options	Per-prompt settings	Every prompt

The View full screen button opens the whole panel, with the list, in a large window.

Anatomy

Anatomy shows the Request Anatomy: a labeled map of every message in the request this prompt sends. It uses your live settings. In a game, it uses that game's values. Outside a game, it uses sample values.

Tab	What it shows
Chips	Your template, with each value as its chip. Dashed chips are parts the app fills in. Select one to open the prompt that writes it.
Preview	The full text of the request. The highlighted parts are text you typed. Select one to open the editor you typed it in.

On a wide screen, a button shows Chips and Preview side by side. When your settings never send this request, Anatomy says so and shows nothing.

System Prompt

The System Prompt is the main text of the prompt. It tells the AI its job and holds the chips for the world, the scene and the story so far. Edit it in the chip editor.

User Message

The User Message is the last message of the request. It usually holds your action or the narration, and the task for this reply. Narration has it only while Thinking is Native. In the other thinking modes, the app writes that message itself.

Messages

Messages holds the extra lines that Narration puts in its history. Each field shows when the app sends it, and has its own Reset.

Message	Sent when
Recap Message	Memory Summaries has condensed older turns. It asks for the story so far, and the summaries answer it.
Now Message	At the end of the recap. It says where things stand now.
Recall Message	Scene Recall brings back an old turn. It marks the scene as the past.
Direction Message	Your action has text in [square brackets], while Thinking is Native. It tells the AI the brackets are your direction as the author.

A field shows only while its feature can send it.

Options

Options holds the settings of one prompt. On a built-in preset they are read-only.

Option	What it does	Shows on
Endpoint	Sends this prompt to an endpoint preset. Use Active Endpoint follows the Endpoints tab. See How to Route a Prompt to Another Endpoint.	Every prompt
Include Attachments	Sends the images attached to your action with this prompt. On for Narration only, by default.	Prompts that send your action, while Image Attachments is on
Max Output	Caps the answer length of this prompt. Off shows Auto and the shipped cap. It does not cap reasoning.	Planning, Director, Character, Storyboard, Choices, Discover Entity, Summaries, Milestone Select, Diary, Scene Tags
Verbatim Turns	Sets how many recent turns go word for word. Older turns go as summaries.	Narration, Planning, Choices, Stat Updates, Location Change, Summaries, while Memory Summaries is on
Native Reasoning	Sets whether this prompt reasons, and how hard. Clear the checkbox and the prompt does no reasoning. Global follows Native Reasoning in Settings → Output. Model Default sends no hint.	Endpoints whose model reasons. Not on Narration with Thinking set to Inline.
Reasoning Budget	Sets the reasoning tokens as a percent of the endpoint's Max Output Tokens. The answer keeps its full length.	Endpoints that take a token budget, such as the desktop engine
Custom Temperature	Sets this prompt's temperature. Off shows the value the prompt sends now.	Every prompt
Custom Repetition Penalty	Sets this prompt's repetition penalty. Off shows the value the prompt sends now.	Every prompt

Reasoning Budget needs Max Output Tokens on the prompt's endpoint. Without it, the slider is off, and a note tells you to set it. A model that always reasons keeps the Native Reasoning checkbox on, with a note.

The Chip Editor

Each prompt editor is a chip editor. A chip is a box in the text. When the app sends the request, it puts a value in each chip, such as the world description, the stats or your action.

• A chip sends text, and no chip sends nothing. The app adds nothing to a prompt that you did not place. Delete the Notes chip, and the AI never reads your notes.
• The bar above the editor lists the chips this prompt can use. Select a chip to put it at the cursor, or drag it into the text.
• Select a placed chip to open Chip Options. The dialog sets what the chip sends. See Chip Options.
• The editor has two tabs. Edit holds the template. Preview shows the text with each chip filled in. Outside a game, a Sample data tag marks the sample values. Preview is off while the prompt has no chips.

The editor has Undo and Redo buttons. Edit full screen opens the editor in a large window. For the chips that carry the persona and names, see Personas for Authors.

Chip Options

Chip Options is a dialog with the settings of one placed chip. It shows only the rows that chip has.

• A prompt chip has two panes. The left pane holds the rows. The right pane, What the AI Reads, shows the text the chip sends with your picks. It includes the Header, Prepend and Append text, in a dimmer shade.
• Hover or focus a row to see what it adds. The text it adds lights up, and the rest dims. A Stat Detail piece lights only its own part of each stat line.
• The pane reads real data where it can. In a game it reads the live scene. In the World Editor it reads the authored world. In Settings it reads sample data.
• Placeholder chips and Tool argument chips have one pane. They show their rows and no What the AI Reads.
• A narrow screen stacks the panes. The rows come first.
• A list of four options or more is a dropdown. A short list is a row of buttons. Hover an option to read what it does.
• Read-only presets show every row, disabled.
• Re-Pick… aims a placeholder chip at another value. Edit Value opens the chip's value for typing.

Row	Chips	What it does
Scope	Location, Entities	Picks which locations or entities the chip lists
Only With	Entities, Persona	Keeps only entities that own Stats, Traits or either
Content	Location, Entities, Persona	Sends Full detail, a Summary, or the Name only
Include	Entities, Persona	Sends each entity's Traits and Stats, each on or off
Stat Detail	Stats, Entities, Persona	Sends Range, Descriptor and Description for each stat. At least one stays checked.
Format	Stats, Traits, Location, Entities, Persona	Shapes the text as Simple, Markdown or XML
Mode	Markdown Guidance, Dictionary	Picks the one form the chip sends, such as Foreground or Background lore
Label, Header, Prepend, Append	Placement fields, below a rule	Header starts a section. Prepend and Append wrap the value. A chip with no value sends none of them.

• Include Stats is on by default for Here and In Scene. It is off for In Sub and Reachable.
• Name content turns off Include and Stat Detail. Only With still works, so the name list holds only the entities that pass.
• Include Stats off turns off Stat Detail.
• A bare chip sends what it sent before. Chips in saved presets and worlds keep working.

⚠️ An older version of Formamorph can't read the new picks. A preset or world that uses Only With, Include or Stat Detail picks shows the raw chip text there, and the AI reads it as text.

How to Send Only Entities With Stats

Use Only With on the Entities chip. The chip then lists only the entities that own stats.

1. In the prompt, select the Entities chip. Chip Options opens.
2. Under Only With, check Stats.
3. Optional: check Traits too. The chip then keeps an entity that owns stats or traits.
4. Read What the AI Reads to check the list.

• The chip shows a filter icon while a pick is on. Hover the icon to read the picks, such as "Only With Stats or Traits". The chip's label does not change.
• An entity owns stats when at least one of its stats is on. The same holds for traits. A switched-off stat or trait doesn't count.
• Include doesn't change who stays. An entity with stats stays even when you clear Include → Stats.
• The Persona chip has the same row. A persona that fails the pick sends what no persona sends: N/A, or nothing when the chip has a Header, Prepend or Append.

How to Get Entity Stats Into a Custom Stat Prompt

The default Stat Updates prompt reads four blocks. A Stat Updates prompt you saved keeps its own chips, so it sends no entity stats until you add the Entities chip. The Stats chip sends no In Scene entity's stats.

Chip	Header	Sends
Persona	Player Character	The persona's traits and stats
Entities	In Scene	Each present entity that owns stats
Stats	Other Stats	World stats that no persona or entity owns
Traits	Other Traits	World traits that no persona or entity owns

1. Open the Prompts tab in Settings, in Advanced mode. Select Stat Updates in a preset of your own.
2. Put the cursor after your Persona chip. Select the Entities chip in the bar.
3. Select the placed chip to open Chip Options.
4. Set Scope to In Scene.
5. Under Only With, check Stats.
6. Set Content to Summary. Under Include, check Traits and Stats. Under Stat Detail, check all three pieces.
7. In Header, type In Scene.
8. In the format section of the prompt, add a line that lists the reply keys: Player Character, Other Stats and each In Scene entity's name.

• An entity's reply key is its name, as the block shows it. A name that clashes with an earlier owner reads as "name (2)".
• Set the Persona chip's Only With, Content and Include rows the same way. Keep its Header as Player Character. A persona with no stats sends no block.
• A world's own Stats prompt follows the same steps. See World Prompts and the Diff Viewer.

World Prompts and the Diff Viewer

A world can bring its own Narration, Choices or Stats prompt. An author writes them under Custom Prompts in the World Editor's Overview tab. See World Editor Overview. A world prompt takes the place of your preset's prompt for that pass.

When you select such a world on the main menu, a notice under its details names the custom prompts, such as "This world uses a custom narration prompt." Select View on the notice to open Custom Prompts.

Custom Prompts compares the world's prompt with the prompt Formamorph ships for that pass. It does not compare with your own preset.

Control	What it does
Changes	Shows one text with the changes marked. Added text is tinted green. Removed text is struck through in red.
Raw	Shows the world's prompt as the author wrote it
Pass tabs	Narration, Choices and Stats. Show when the world rewrites more than one pass.

Chips show as their raw text, such as <WORLD DESCRIPTION>. To play the world with your own prompts, clear Use this world's prompt under the world's buttons on the main menu. See The World Dialog.

Related

• 🧰 Tools: the functions the AI can call, and which preset turns them on
• ⚙️ Settings: the features that add or remove prompts, and the text endpoints
• 🧠 Story Memory: what Summaries, Milestone Select and Diary keep