Claude Style Illustrationv1.0.2
Use this Skill to make article illustrations.
An illustration skill for Claude Fable 5: one idea becomes a quiet paper-cut metaphor in flat muted color, off-white cut shapes and wobbly black felt-tip lines.
Picked most often by
Updated Aug 25, 2026
Preview gallery
FALSimilar Skills
View all
What you get
- One picture per concept carrying one metaphor of two to four elements, square by default or 16:9, 9:16, 4:3 or 3:4
- Six style reference images and a fill-in prompt frame that hold a series to the same three-color look
- A FAL generator that records each paid request first, and Python scripts that flatten the palette, check the ratio and validate a batch
Good fit
- Hero images and thumbnails for blog posts, newsletters and product pages
- Abstract themes like privacy, growth or teamwork that need one plain symbol
- Post series that should share a single look
Not for
- Photoreal, 3D, textured or highly detailed art
- Infographics, UI mockups or text-heavy images
- Full rooms, streets or landscapes, realistic anatomy or comic-style panels
About this Skill
Writers and product teams who want a warm, human lead picture name a topic; the agent boils it down to a noun and a verb, such as privacy plus protect, and picks one anchor object, adding a simple looped-finger hand only when it makes the action clearer. Two to four bundled style references guide a strict single-picture prompt. Each picture is its own paid request, saved before polling, then flattened by a Python script to three colors in any of five ratios, from square 1:1 to wide 16:9.
How it works
- Give a topic; mention an action, object or ratio only when it matters.
- The agent reduces it to a noun and a verb and picks one anchor object.
- It fills the single-picture frame and attaches two to four references showing line, shapes and hands.
- Each concept gets one generation, its request saved to a task record before polling.
- The raw file is kept; a flattened three-color copy is cut to the exact ratio and checked.
Checked before delivery
- Each picture is one standalone image with one idea in two to four elements, never a grid, collage or set of panels
- After flattening only a flat muted background, warm off-white cut shapes and loose near-black lines remain
- No unrequested text, frame, date, interface, logo or watermark, and nothing photographic, 3D, glossy or vector-perfect
- check_output.py confirms the file and ratio, and exactly one paid request is made per approved picture
What you need before you start
- You bring
- The topic or idea the picture has to get across · Optional: an action, object or metaphor you already have in mind · Optional: aspect ratio (1:1, 3:4, 4:3, 9:16 or 16:9; 1:1 by default) · Optional: how many pictures, one concept per picture
- Dependencies
- Node.js 20 or later, to run the bundled generator · Python 3 with Pillow, to run the flatten and check scripts · Optional: your FAL account with FAL_KEY set, for the bundled generator; FAL bills each request
- Outside costs
- FAL_KEY
- Last tested
- v1.0.2 — Claude Code / Codex / Cursor
User reviews
Loading reviews…
FAQ
Can it make a set of pictures in one go?
Yes, with one generation and one file per concept. A gallery can be assembled afterwards as a review sheet, but it never replaces the single files.
What does the flattening step do?
Image models add soft tones. flatten_style.py trims the picture around its center to the exact ratio and maps every pixel to near-black ink, warm off-white or your chosen background, so each picture in a series matches. The raw file stays untouched.
Is a FAL key required?
Only for the bundled generator, and each request it sends uses paid FAL credits. If your agent has its own image tool and you ask for it, the same prompt frame applies and the Python scripts still run.