A Claude Code skill that generates hand-drawn, marker-doodle style flowchart images for YouTube videos (or any explainer) on any topic — in one consistent, channel-ready aesthetic.
You call the skill, give it a topic, and it designs the slide and generates the
image automatically using OpenAI's image API (gpt-image-1).
The example above was generated by the skill from the prompt: "a fully automated pipeline that scrapes Twitter/X profiles, analyzes them, finds viral posts, has Claude build the project, auto-records a video, and publishes to YouTube."
Every image comes out in the same look, so a whole video series stays on-brand:
- Off-white paper background, felt-tip marker texture
- Big hand-lettered title with a loose purple scribble underline
- A row of 5–6 rounded pastel cards, each with a real hand-drawn icon and a short caption above it
- Diamond connectors between cards, start label (left) → end label (right)
- Dashed question-mark callouts hanging under each step
- A dashed "recipe" summary strip at the bottom
- Palette: peach → lavender → light blue → mint → pink → yellow on off-white
The aesthetic is locked in a fixed STYLE BLOCK inside the skill; only the CONTENT BLOCK changes per topic. That's what keeps every slide consistent.
Requirements
- Claude Code
- Python 3.8+ (standard library only — no pip installs needed)
- An OpenAI API key with image generation access
Quick install
git clone https://github.com/MomoXplore/handdrawn-presentation-skill.git
cd handdrawn-presentation-skill
./install.shThis copies the skill to ~/.claude/skills/handdrawn-presentation/.
Manual install
Copy the skills/handdrawn-presentation/ folder into ~/.claude/skills/.
-
Set your OpenAI API key (the skill reads it from the environment — it is never stored or committed):
export OPENAI_API_KEY=sk-... -
In Claude Code, run the skill:
/handdrawn-presentationor just say "make a presentation for my video". You can include the topic inline:
/handdrawn-presentation a flowchart about launching a SaaS side project -
Claude asks for / refines the topic, designs 5–6 stages, generates the image, and shows it to you. The PNG saves to
./handdrawn/by default.
export OPENAI_API_KEY=sk-...
python3 skills/handdrawn-presentation/scripts/generate.py \
--prompt-file my_prompt.txt \
--out out.png \
--size 1536x1024 --quality highskills/handdrawn-presentation/SKILL.md— the workflow + the locked style recipe and the per-topic content template Claude fills in.skills/handdrawn-presentation/scripts/generate.py— a dependency-free Python script that POSTs the prompt to OpenAI's image API and saves the PNG. Reads the key fromOPENAI_API_KEY.
- The script reads your key from the
OPENAI_API_KEYenvironment variable only. No key is read from, written to, or committed in this repo. .gitignoreexcludes.env, generated images, and scratch prompt files.- Never paste your API key into a chat or commit it. If you do, rotate it at https://platform.openai.com/api-keys.
MIT — see LICENSE.
