A real run of this skill. The input is a synthetic portrait, not a customer photo.

Ready to use

Doodle Avatar

One hand-drawn doodle avatar from your photo

Use this skill

Your photo: Required
Output: Single square 1:1
Category: avatars
Cost: 1 credit

portrait
avatar
marker

What it does

What you get back

Preserves your recognizable hairstyle, face shape, expression, skin tone, and accessories, then redraws them in cute naive marker-and-ink cartoon style: bold clean outlines, simplified features, flat cheerful colors, and a clean warm-white background. No photorealism.

When the agent picks it

Use when the user wants a single hand-drawn doodle avatar from one photo, or simply says "doodle me" / "make me a doodle" with no other detail. This is the default choice when the request is a plain portrait and nothing else fits.

The full skill file

Every skill is one SKILL.md — this is the exact file the agent reads.

src/mastra/skills/doodle-avatar/SKILL.md

Doodle Avatar

Turn one uploaded photo into a single naive doodle fashion-chibi avatar.

When this is the right skill

Pick this skill when the user wants one illustrated portrait of the person in their photo. It is the fallback for vague requests such as “doodle me” or “turn this into a cartoon”.

Choose a different skill instead when the user asks for:

multiple panels, expressions, or a grid → doodle-collage
head-to-toe or action poses → full-body-collage
artwork with no photo at all → surprise-me

What to draw

Read the photo and keep the subject recognizable: hairstyle and hair color, face shape, expression, skin tone, glasses, and any distinctive accessories. Then redraw all of it as illustration — never a filtered photograph.

House style (see references/style-guide.md for the full spec):

bold, clean, confident outlines
simplified facial features, no rendered detail
flat cheerful colors with minimal shading
playful, slightly exaggerated proportions
clean white or warm-white background

How to run it

Confirm a photo is attached. If none is, ask the user to attach one — do not call the tool without a photo. If they would rather not upload anything, offer surprise-me instead.
Call generateDoodle with skill: "normal" and the uploaded photo as imageUrl.
If the user also attached a style reference (their message contains a line like Reference image: <url>), pass it as refImageUrl.
Report the result briefly. The app displays the image itself, so never paste the URL back into the conversation.

Following up

After a successful generation, offer one concrete next step — a collage of several expressions, full-body action poses, or a refinement such as thicker outlines or a warmer paper tone.