Draw logo

Draw

Organization
with-pebbly
draw

Draw a sprite from silhouette to finished pixels, in the order that catches mistakes while they are still cheap — block in, check the silhouette, shade, outline, verify. Use for the main body of any drawing task.

Overview

Publisherwith-pebbly
Repositoryaseprite-ai-artist
Skill namedraw
Stars
456
Forks
28
Bundled files
Instructions only
LicenseMIT
Links
  • Markdown instructions

    A SKILL.md file the model loads on demand, so it only costs tokens when a request actually matches.

  • Works with any LLM

    AI skills are plain Markdown, not provider-specific code, so this works with GPT, Claude, Gemini, Grok, or a local model.

  • Self-contained

    Everything the model needs lives in the instructions — no extra files to sync.

  • Open source

    Published by with-pebbly on GitHub. Read the source before you install it.

Installation

Install the Draw AI skill in TypingMind to use it with any LLM, or drop it into another agent that reads SKILL.md.

1

Install in TypingMind

TypingMind installs a skill straight from its GitHub folder — it reads SKILL.md, bundles the resource files, and stores the result locally.

  1. Open the app and go to Plugins → Skills.
  2. Choose "Install from GitHub".
  3. Paste the skill folder URL below and confirm.
    https://github.com/with-pebbly/aseprite-ai-artist/tree/main/skills/draw
  4. Enable the skill in any chat where you want it available.
Plugins → Skills → Add skill → From GitHub URL, then paste the folder URL and press Continue.
2

Install in another agent

Any agent that reads the Agent Skills format can use this skill — copy the folder into that agent's skills directory.

Claude Code — .claude/skills
git clone --depth 1 https://github.com/with-pebbly/aseprite-ai-artist.git /tmp/aseprite-ai-artist
mkdir -p .claude/skills
cp -r /tmp/aseprite-ai-artist/skills/draw .claude/skills/draw
Restart Claude Code after copying so it picks up the new skill.

Use it in TypingMind

Enable Draw in any TypingMind chat and the model takes it from there. Its name and description sit in the system prompt, and the moment a request matches, the model loads the full instructions itself — you never invoke it by hand, and it costs no tokens until it is actually used.

The model loads Draw on its own as soon as a request matches it.

Works with any AI model

AI skills are plain Markdown instructions rather than provider-specific code, so Draw is not tied to the model it was written for. Install it once in TypingMind and use it with GPT-5, Claude, Gemini, Grok, DeepSeek, Mistral, Llama, or a local model you run yourself — all on your own API keys.

  • Loaded only when it is needed

    The system prompt carries just the name and description. The instructions are fetched on the first matching request, so an idle skill costs nothing.

  • Switch models mid-chat

    Because the skill is instructions rather than code, changing model does not break it — the next model reads the same SKILL.md.

Skill instructions

This is the SKILL.md content the model loads. Read it before installing — a skill is instructions your model will follow.

Draw a sprite

The order matters more than the technique. Detail added before the form is right just makes the wrongness harder to see and more expensive to fix.

Procedure

1. Look before you draw

preflight, then sprite_info. You need the real layer names, frame count and palette. Do not assume them — a wrong layer name means drawing into the user's finished art.

If aseprite:concept produced a PixelSpec, it is the plan: its landmarks and bounding boxes are where the shapes go, its palette mapping is which colour each mass gets, its deviations are what you leave out. When a reference layer exists (reference op="list"), you are drawing over it — the steps below are the same, and step 3 gains a comparison.

Then read the subject rules for what you are drawing — the table in skill://studio under Subject rules names them. A hand, a face, a horse, an isometric crate or a waterfall each has its own file with size budgets and ```grid templates. A template is a starting point you transcribe with draw op grid, mapping each legend role onto the sprite's palette; adapt it to the pose rather than inventing the shape from nothing.

2. Block in the silhouette

One flat colour, no detail. Target the base layer. Everything in one draw call.

For anything up to about 32×32, write it as a grid: one character per pixel, so you see the whole silhouette while you write it instead of finding out afterwards what a stack of ellipses and rects added up to.

draw layer="base" label="block in slime" ops=[
  { kind:"grid", x:8, y:10, legend:{ "O":"#5f574f" }, rows:[
    "....OOOO....",
    "..OOOOOOOO..",
    ".OOOOOOOOOO.",
    "OOOOOOOOOOOO",
    "OOOOOOOOOOOO",
    ".OOOOOOOOOO." ] }
]

Every row must be the same width, . is transparent, and transparent cells erase what is under them (transparent:"skip" to stamp over existing art instead). Bigger sprites: one grid per body part or per layer, each at its own x/y — long rows of identical characters are where a model miscounts, and a ragged row is refused with its row number rather than silently shifted.

Shapes are still the right tool for big regular forms — a sky gradient, a floor, a 40-pixel circle:

draw layer="base" label="block in knight" ops=[
  { kind:"ellipse", rect:{x:12,y:4,width:8,height:8}, color:"#5f574f", fill:"#5f574f" },
  { kind:"rect",    rect:{x:11,y:12,width:10,height:10}, color:"#5f574f", fill:"#5f574f" },
  …
]

Batch aggressively. One call is one undo step for the user; forty calls are forty. See rules://03-silhouette-and-form.

3. Check the silhouette immediately

look op="preview"

Ask yourself the only question that matters here: is it recognisable as one flat shape? If not, fix it now. Shading a bad silhouette is wasted work.

Watch for: symmetry that reads as a statue, tangents where an arm fuses into the torso, limbs all the same thickness.

With a reference, preview blends the half-transparent reference into the art — use look op="compare" instead: reference left, art right. Name the three to five largest mismatches (silhouette, proportion, pose, where the colour masses sit), fix only those in one draw call, compare again. Stop when what is left is a deviation the PixelSpec lists. This loop repeats after materials (step 4) and after shading (step 5): the reference is a check on every stage, not only the first.

4. Separate the materials

Replace regions of the blockout with each material's base colour — skin, metal, leather, cloth. Still flat. Still one draw call.

5. Shade

See aseprite:shade. One shadow step, look, one light step, look. Stop there unless the sprite is 32px+ and genuinely needs more.

6. Outline

Pick one style from rules://04-outlines-and-edges and apply it consistently. Selective outlining — outside only — is usually right. transform op outline does the mechanical part: side="outside" grows the shape by a pixel, side="inside" recolours its edge and keeps the size; diagonals=true fills the corner pixels for square corners, off leaves them cut and softer. Hand-place where you want it broken.

6b. Text, if the art has words

draw op kind text lays out a string from a bitmap font and draws it in the same batch as everything else. Pass measureOnly: true with only text ops first to get the ink bounds back without touching the sprite — that is how you centre a label or size a panel around it before committing to a position. See TOOLS.md for the font format.

7. Verify precisely

look op="ascii"

The text grid is where you catch the pixel one row too low and the line run of 3 in a sequence of 2s. A preview cannot show you those.

To fix what it shows, edit the grid itself: look op="ascii" layer="base" rulers=false region=… returns bare rows plus origin; change the cells that are wrong and send the rows back as draw kind grid at that origin, with the legend look gave you. Only the region you send is touched. A glyph is a palette index and means the same colour in every region you read.

8. Validate and report

validate

Fix every error. Report warnings you chose not to fix, with the reason.

Batching rules

  • Ops run in array order, so paint fills before outlines and outlines before highlights.
  • Leave paletteLock on. When the result says a colour moved with ΔE > 12, the palette has no colour for what you asked — say so rather than turning the lock off.
  • Use label to describe the intent; it becomes the user's undo entry.

Related

rules://00-core-principles, rules://02-shading-and-light, rules://03-silhouette-and-form, rules://04-outlines-and-edges, rules://10-lines-and-curves, rules://11-clusters-and-noise, and the subject files listed in skill://studio under Subject rules.

Frequently asked questions

What does the Draw AI skill do?

Draw a sprite from silhouette to finished pixels, in the order that catches mistakes while they are still cheap — block in, check the silhouette, shade, outline, verify. Use for the main body of any drawing task.

Why use Draw on TypingMind?

Because you install it once and use it with any model. Draw is plain Markdown rather than provider-specific code, so the same skill runs on GPT-5, Claude, Gemini, Grok, or a local model — and you can switch model mid-chat without it breaking. TypingMind runs on your own API keys, so you pay providers directly instead of a per-seat subscription, and your skills and chats stay in your own storage.

How do I install Draw in TypingMind?

Open Plugins → Skills → Install from GitHub in TypingMind and paste https://github.com/with-pebbly/aseprite-ai-artist/tree/main/skills/draw. TypingMind reads its SKILL.md and installs it as a skill you can enable per chat.

Which AI models can use Draw?

Any model you connect in TypingMind. AI skills are plain Markdown instructions rather than provider-specific code, so GPT, Claude, Gemini, Grok, and local models can all load this skill when a request matches it.

How many AI models can I use with Draw?

As many as you like. As long as a model supports skills, you can use Draw with it — GPT, Claude, Gemini, Grok, DeepSeek, Mistral, Llama and more — all on TypingMind with your own API keys.

Is the Draw AI skill free?

Yes. It is published on GitHub by with-pebbly under the MIT license. You only pay your own AI provider for the tokens you use.

What are AI skills?

An AI skill is a reusable instruction bundle that teaches an AI model how to do one specific task. It follows the open Agent Skills format: a SKILL.md file with a name and description, plus any scripts, templates or reference files the model may need. The model reads the instructions only when your request matches the skill, so an installed skill costs nothing until it is used.

How are AI skills different from plugins or MCP servers?

A plugin or MCP server gives a model new tools to call — code that runs somewhere and returns a result. An AI skill gives the model knowledge and process instead: how to approach a task, which steps to follow, what good output looks like. Skills are plain Markdown, so they need no server, no API key and no runtime, and they work with any model.

View all

Set up your own AI workspace now

Get notified about new features and future giveaways by subscribing to our newsletter 👇