tripo-prompt
Prepares everything a Tripo (Tripo3D, Tripo Studio) 3D model needs before the user generates it - the input picture spec (single view or front/left/back/right views, with a prompt to make them), the text prompt and negative prompt for text-to-3D, the texture prompt, and a Tripo card of settings (generator and AI model, polycount, topology, texture, rig, export) - and reads the result to say what to change. Use when the user asks for a Tripo prompt, a 3D model prompt, pictures or views for image-to-3D or multiview-to-3D, Tripo settings for a prop, building, vehicle, creature or character, a retexture, a rig, or reports or shows how a Tripo model came out. Prompts and settings only: this skill generates no art and no models.
Safety scan
A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.
Badge for your README
Markdown
[](/listings/sabdakan--tripo-prompt-skill)HTML
<a href="/listings/sabdakan--tripo-prompt-skill"><img src="/api/listings/sabdakan--tripo-prompt-skill/badge.svg" alt="findsafeskills safety rating" /></a>Install
Install instructions from the README
As written by the repo's author. findsafeskills has not verified these steps.
macOS or Linux:
git clone https://github.com/Sabdakan/tripo-prompt-skill.git ~/.claude/skills/tripo-prompt
Windows (PowerShell):
git clone https://github.com/Sabdakan/tripo-prompt-skill.git "$env:USERPROFILE\.claude\skills\tripo-prompt"
**Update** to the newest version at any time:
git -C ~/.claude/skills/tripo-prompt pull
(on Windows: `git -C "$env:USERPROFILE\.claude\skills\tripo-prompt" pull`)
What was scanned
The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.
SKILL.md · Manifest · 105 lines
- ---
- name: tripo-prompt
- description: Prepares everything a Tripo (Tripo3D, Tripo Studio) 3D model needs before the user generates it - the input picture spec (single view or front/left/back/right views, with a prompt to make them), the text prompt and negative prompt for text-to-3D, the texture prompt, and a Tripo card of settings (generator and AI model, polycount, topology, texture, rig, export) - and reads the result to say what to change. Use when the user asks for a Tripo prompt, a 3D model prompt, pictures or views for image-to-3D or multiview-to-3D, Tripo settings for a prop, building, vehicle, creature or character, a retexture, a rig, or reports or shows how a Tripo model came out. Prompts and settings only: this skill generates no art and no models.
- ---
- # Tripo prompt writer
- If PROJECT-NOTES.md exists beside this file, read it first: its project rules win over this skill's defaults.
- The user makes every model themselves in Tripo. This skill prepares the inputs and settings and reads the results. In Tripo the picture is the bigger half of the prompt: a clean input picture does more than any wording.
- ## Pick the job
- | The user asks for | Mode | Read |
- |---|---|---|
- | Anything that moves (creatures, a windmill's sails, a door): the motion card, then what it asks of pictures and settings | Moving asset | [MOVING.md](MOVING.md) |
- | A model that starts as 2D art (the usual case): design, then model reference, then views, then 3D | The 2D-to-3D chain | [FROM-2D.md](FROM-2D.md) |
- | A model from an approved design picture | Image to 3D | [INPUT-IMAGES.md](INPUT-IMAGES.md), [SETTINGS.md](SETTINGS.md) |
- | A model seen from every side (a turning camera or map; creatures, characters, buildings, hero objects) | Multiview to 3D | [INPUT-IMAGES.md](INPUT-IMAGES.md), "Views" |
- | An approved model to file for the team | The shared set | [FROM-2D.md](FROM-2D.md), "The shared set" |
- | The pictures themselves do not exist yet | Picture brief first | [INPUT-IMAGES.md](INPUT-IMAGES.md), "Writing the picture prompt" |
- | A simple, common object with no design (a rock, a barrel, a crate) | Text to 3D | [TEXT.md](TEXT.md) |
- | New paint on a model whose shape is right | Retexture | [TEXT.md](TEXT.md), "Texture prompt" |
- | Skeleton, animation, parts, low-poly, export | After generation | [SETTINGS.md](SETTINGS.md), "After generation" |
- | "Which model, what polycount, can I redo it, what does it cost" | Settings and cost | [SETTINGS.md](SETTINGS.md) |
- | They say or show how a model came out | Result | [FEEDBACK.md](FEEDBACK.md) |
- | The first model ever for a project | Setup, then the job | [FEEDBACK.md](FEEDBACK.md), "First time for a project" |
- Worked answers of each kind are in [EXAMPLES.md](EXAMPLES.md).
- ## Workflow
- 1. **Brief.** Five things: the project; the asset and its kind (prop, building, creature, character); how it is seen in the game (size on screen, the camera's angle, and whether the player can turn the camera or the map); whether it moves; what exists (design pictures, views, an earlier model). Ask at most two questions; show other choices on the ASSUMED line.
- 2. **Read the project's files**, lessons first: `docs/tripo/prompt-log.md` (Lessons), the project's 3D budget and pipeline docs, and PROJECT-NOTES.md when it exists.
- 3. **Steady or moving?** A moving asset gets a motion card first (MOVING.md), and the answer carries a MOTION line.
- 4. **Choose the mode** from the table. Prefer pictures over text whenever a design exists, and several views over one whenever the back or the underside matters.
- 5. **Pictures.** Check the user's pictures against INPUT-IMAGES.md and say what to fix, or write the prompts for making them. A design picture is not a model reference: turn it into one with an edit (FROM-2D.md), keeping the design. Views come in this order: Tripo's own Generate Multi-Views from the model reference; if they drift, Dreamina's **Angles** tool (new camera angles of the same subject from one reference picture); if that drifts too, one edit prompt per view (INPUT-IMAGES.md, "Views").
- 6. **Card.** Fill the Tripo card from SETTINGS.md and the project's budget. Name the tab and the AI model as the screen shows them, keep the polycount inside that model's limit (table below), and quote the price on the button. The project's own numbers win over this skill's defaults.
- 7. **Check** (list below), then answer in the output format.
- 8. **Learn.** On a result, follow FEEDBACK.md: fix the input before the prompt, one change per try, the try logged.
- 9. **Share.** When the user approves the model, list its shared set (FROM-2D.md) and where it goes, so the rest of the team works from the same pictures.
- ## Tripo's generators, as its screen showed them (20.11.2026)
- | Tab | AI Model | Topology | Polycount | Generate button | Texture |
- |---|---|---|---|---|---|
- | Smart Mesh | P1.0 Fast | triangles only | 500 to 20,000 (default 5,000) | 35 credits | none on this panel: a separate Texture step afterwards |
- | Smart Mesh | P2.0 | quads | 500 to 25,000 (default 5,000) | 100 credits | none on this panel: a separate Texture step afterwards |
- | HD Model | H3.1 | quads or triangles | 500 to 2,000,000 | 55 credits with texture at 4K, PBR and remove lighting | in its Geometry & Texture settings |
- - **Smart Mesh is the game asset generator.** P1.0 Fast for anything steady or made of rigid parts; P2.0 for a model that bends. HD Model (H3.1) only for close-up renders, printing, or a detailed source to reduce.
- - **Number of Generation: 1 / 2 / 4** (members) gives that many tries at **one** polycount. A spread over several counts is one paid run per count.
- - Prices and limits change. The button is the truth: read it, and record anything new in [MODEL-NOTES.md](MODEL-NOTES.md).
- ## Standing rules
- - One object per model. A scene, a group or a character holding a separate prop is several models.
- - Pictures the user did not make are for reading in words; never suggest uploading art the user does not own.
- - No names of artists, studios, films or other games in any prompt.
- - A fact about Tripo is "per the docs" until the user confirms it on their own Tripo screen; confirmed facts go in [MODEL-NOTES.md](MODEL-NOTES.md).
- - Follow the project's rules for generated models (disclosure, what may ship, asset logs).
- - When the structure is wrong, a new try comes before a longer prompt: results vary between tries.
- ## Output format
- ```
- ## <asset> - <project> - <Text to 3D | Image to 3D | Multiview to 3D | Retexture>
- TRIPO CARD Generator: <Smart Mesh, P1.0 Fast | Smart Mesh, P2.0 | HD Model, H3.1> · Polycount: <inside that model's limit> · Topology: <triangles | quads> · Number of Generation: <1 | 2 | 4, tries at this one count>
- Texture: <Smart Mesh: the later Texture step, standard or HD, PBR on/off | H3.1: in Geometry & Texture, 2K/4K/8K, PBR on/off, remove lighting on/off>
- After: <Edit Mesh / parts / rig / nothing> · Export: <format, pivot, bottom> · A try costs: <credits on the button, when known>
- MOTION <steady | rigid parts: which, turning where | bends: where; who animates it>
- PICTURES <which picture goes in which slot, each a separate file>, or "none (text)"
- ASSUMED <choices made for the user>, or leave the line out
- PROMPT <text mode and retexture only: one code block to copy whole>
- NEGATIVE <text mode only: a few words, or leave out>
- LOOK AT <a full turn first, then from the game's camera; then three to five things for this asset>
- IF IT COMES OUT WRONG
- - <likely problem> -> <the one thing to change>
- ```
- When pictures must be made first, the answer is the picture brief (INPUT-IMAGES.md) followed by the card that will be used once they exist.
- ## Check before answering
- - [ ] The picture going into Tripo is a model reference, not the design picture; it was made from the approved design by an edit, and card text is pasted unchanged.
- - [ ] One object; nothing in the picture or prompt that is not part of the model (ground, shadows, scenery, effects).
- - [ ] The picture rules hold: whole object, centred, plain background, even light, no cast shadow, neutral view, parts clear of each other.
- - [ ] Creatures and characters are in a neutral, symmetric pose with every limb visible; counts are stated (four legs, one tail).
- - [ ] A model the player can see from every side has four views, and none of them is treated as "the back".
- - [ ] Views are of the same object, pose, scale and light: front, left, right, back. Each view is its own picture file, never several in one picture. They come from Generate Multi-Views first, then Dreamina's Angles tool, then one edit prompt per view. A top view is a check picture, never an input.
- - [ ] A moving asset has its motion card: parts that move are drawn apart with their joints, a model that bends is in a neutral pose on Smart Mesh, P2.0 (quads), and a steady one pays for neither.
- - [ ] The generator fits the use and is named in full: Smart Mesh, P1.0 Fast or P2.0, for a game asset at the polycount it needs; HD Model, H3.1, only for close-up renders, printing or a source to reduce.
- - [ ] The polycount is inside the model's limit (P1.0 Fast 20,000; P2.0 25,000; H3.1 2,000,000). Number of Generation is tries at one count; several counts are several paid runs.
- - [ ] Text prompts read object, style, material, form, proportions: short keywords, no mood, light or camera words.
- - [ ] Thin parts are thick enough to survive (legs, tails, ears, straps, blades), or are named as a risk.
- - [ ] The card's polycount and texture come from the project's budget and from what its engine material actually uses (no PBR or HD texture the game ignores).
- - [ ] What happens after generation is on the card and matches the project's pipeline (who rigs, where low-poly happens).
- - [ ] Every Lesson in the project's prompt-log is applied, and the standing rules hold.
- ## Files
- FROM-2D.md · MOVING.md · INPUT-IMAGES.md · TEXT.md · SETTINGS.md · FEEDBACK.md · MODEL-NOTES.md · EXAMPLES.md · PROJECT-NOTES.md (the user's own projects; it may not exist)
- [MANUAL.md](MANUAL.md) is the user's guide to this skill. Read it only when they ask how to use the skill; keep it true when the skill changes.
- Works with the `dreamina-prompt` skill when installed: this skill says what the picture must be, that skill writes the prompt for it.
README.md · README · 41 lines
- # tripo-prompt
- A Claude skill for Tripo Studio (AI image-to-3D and text-to-3D): the input pictures and their checks, the prompts, the settings card (generator, polycount, texture, export), moving versus steady models, and reading the result to fix the next try. Made for game assets drawn first in 2D.
- It writes prompts and settings only; you run them in the tool yourself.
- ## Install in Claude Code
- macOS or Linux:
- ```
- git clone https://github.com/Sabdakan/tripo-prompt-skill.git ~/.claude/skills/tripo-prompt
- ```
- Windows (PowerShell):
- ```
- git clone https://github.com/Sabdakan/tripo-prompt-skill.git "$env:USERPROFILE\.claude\skills\tripo-prompt"
- ```
- **Update** to the newest version at any time:
- ```
- git -C ~/.claude/skills/tripo-prompt pull
- ```
- (on Windows: `git -C "$env:USERPROFILE\.claude\skills\tripo-prompt" pull`)
- ## Install in the Claude app
- Download [tripo-prompt.zip](https://github.com/Sabdakan/tripo-prompt-skill/releases/latest/download/tripo-prompt.zip) and add it in the app's Skills settings. The link always gives the newest version: download it again and replace the skill to update.
- ## Your own project's rules
- Write them in a file named `PROJECT-NOTES.md` beside `SKILL.md`: your style, your characters, sizes, who makes what. The skill reads it first, and its rules win over the skill's defaults. Git ignores the file, so updates never touch it.
- ## Start here
- Read [MANUAL.md](MANUAL.md): how to talk to the skill and what each answer contains.
- Works well with [dreamina-prompt](https://github.com/Sabdakan/dreamina-prompt-skill).
Ratings and comments
Sign in to rate this skill.
Sign inFree to use. If it’s helped, pass it on.