Your selfie-to-makeup stylist: pick a look, preview it, then learn exactly where to apply it.
"I'm going out tonight, but I have no idea what makeup suits my face."
"I want to buy makeup for my girlfriend, but I don't know what style would flatter her."
"The final makeup looks good, but where do I actually put the blush, liner, and lip tint?"
Makeup Skill turns one portrait into a visual makeup plan. It recommends flattering directions, creates realistic preview options, and builds a step-by-step tutorial with where to apply on the left and expected result on the right.
It is designed to preserve the person's real face. No face reshaping, no identity drift, no beauty-filter fantasy. Just makeup color, placement, texture, and technique.
- Recommends 3 to 5 makeup directions for your face, outfit, occasion, or mood.
- Generates a labeled comparison grid so you can choose the vibe quickly.
- Creates a realistic final preview of the selected look.
- Builds a complete visual tutorial: each step shows the application area and the expected result.
- Helps correct in-progress makeup when something feels too heavy, old-fashioned, uneven, or unlike you.
- Uses an experimental identity guard to catch previews that stop looking like the original person.
Best results come from GPT Image 2 when the host agent provides it. If not, the skill falls back to the agent's available image generator.
- "What makeup should I wear for a summer date?"
- "Recommend a soft party look for this selfie."
- "I want something fresh, not mature or heavy."
- "Generate four preview options."
- "Make a full makeup breakdown for the first look."
- "This eyeliner looks too strong. How should I fix it?"
- Parties, dates, weddings, graduation photos, travel days, work events, and everyday looks.
- People who want a visual guide instead of vague product names.
- Gift planning when you want to understand what colors or textures suit someone.
- Makeup learners who need clear placement, not just "apply blush to cheeks."
Implementation details, CLI usage, provider setup, output files, and debugging notes live in docs/technical-readme.md.