rob-janoff-design
Health Pass
- License — License: MIT
- Description — Repository has a description
- Active repo — Last push 0 days ago
- Community trust — 20 GitHub stars
Code Pass
- Code scan — Scanned 12 files during light audit, no dangerous patterns found
Permissions Pass
- Permissions — No dangerous permissions requested
No AI report is available for this listing yet.
Agent Skill for logo design with Rob Janoff's method from "Taking a Bite out of the Apple": design, critique and refresh logos as three one-ink SVG concepts built from a golden series, measured and rendered at 256, 48 and 16 px before they are shown.
rob-janoff-design
An Agent Skill that makes an AI agent design logos, icons and monograms with Rob Janoff's methodology from Taking a Bite out of the Apple:
- Do your homework: business, audience, USP.
- Abstract the shape: silhouette, cross-section, scale relation.
- Add "a wink and a smile".
- Draw three one-ink SVG concepts as vectorized silhouettes, traced from the subject's real contour, and wait for the client's pick; then test the chosen one in color, solid monochrome, at 256, 48 and 16 px and in context.
- Close with a self-critique that says where the logo or the brand identity could be renewed.
Install
Paste this prompt into your agent (Claude Code, Codex, Gemini CLI, GitHub Copilot, Cursor or any agent that supports Agent Skills):
Install the rob-janoff-design skill from https://github.com/carrapido/rob-janoff-design, refer to the repo's AGENTS.md for instructions.
Install with a CLI instead
| Agent | Command |
|---|---|
| Claude Code, Codex, Cursor, Gemini CLI, GitHub Copilot, OpenCode and 70+ more | npx skills add carrapido/rob-janoff-design |
| Claude Code (plugin) | /plugin marketplace add carrapido/rob-janoff-design, then /plugin install rob-janoff-design@rob-janoff-design |
| Gemini CLI | gemini skills install https://github.com/carrapido/rob-janoff-design.git --path skills/rob-janoff-design |
Manual install
Copy skills/rob-janoff-design/ into your agent's skills directory:
| Agent | Personal | Project |
|---|---|---|
| Claude Code | ~/.claude/skills/ |
.claude/skills/ |
| Codex | ~/.agents/skills/ |
.agents/skills/ |
| Gemini CLI | ~/.gemini/skills/ |
.gemini/skills/ |
| GitHub Copilot | ~/.copilot/skills/ |
.github/skills/ |
| Cursor | ~/.cursor/skills/ |
.agents/skills/ |
What changes
Before
|
After
|
From a sketch to three concepts
A second example, starting from a hand-drawn sketch (@marsanwishito): a melancholic girl with a side fringe over one eye and a red heart on her chest, tied by a red thread to a poster. The brief was a T-shirt logo: vectorize her head and hair, imagine how the hair falls behind her head, then design. Each concept pulls a different detail out of the drawing.
|
Reference The sketch traced in vectors, keeping the pencil's width. |
Three details, three concepts
All three start from the same one-ink head: the hair is the mass and the face is the hole. |
|
|
A · Silhouette and cut The fringe becomes a straight cut that leaves one eye. At 16 px the head and the cut still read. |
B · Negative-space wink The heart on her chest becomes her hair: an upside-down heart with the face as its hole. Turn it 180° and the heart appears. |
C · Graphic continuum One line runs from the hair to the chin and turns into the red thread, ending in her heart. At 16 px the line breaks up, so an icon needs a simpler cut. |
Usage
Ask your agent in plain words. The skill picks one of three modes:
| Mode | Example request | What you get |
|---|---|---|
| Design | "Design the logo for a specialty coffee shop in Oaxaca" | Three one-ink concepts, then the full preview of the one you pick and a self-critique of where to renew it |
| Critique | "Critique our logo: logo.svg" | A verdict with evidence (renders, on-screen sizes) against Janoff's golden rules, nothing redrawn |
| Refresh | "Refresh our logo, people already know it: logo.svg" | Rebuild, refine and rethink-one-element options beside the original, like the cat above |
Evals
evals/ holds one case per mode (design, critique and refresh), one that takes a design through to its closing self-critique, and visual cases that judge the logo itself, for claude plugin eval. Each case runs with and without the skill, and graders score it: the commands the agent ran, the files it did or did not write, and an LLM judge for the message it ends with. The agent writes SVGs and runs Node, so the operator grants those tools:
claude plugin eval . --allow-tools Write Edit Bash "Bash(node *)" "Bash(mkdir *)"
Every run is a full agent session billed to your account; add --runs 1 and --max-cost-usd <usd> to keep a check cheap. Results go to evals/results/, which git ignores.
Repository layout
.
├── .claude-plugin/
│ ├── marketplace.json # Claude Code plugin marketplace
│ └── plugin.json # Plugin manifest
├── .github/workflows/test.yml # CI: npm test and the type check on Linux, macOS and Windows
├── AGENTS.md # Install steps the agent follows
├── evals/ # claude plugin eval cases: each mode, the self-critique, the visual judges
├── tests/ # Tests for the skill's scripts: npm test
└── skills/
└── rob-janoff-design/
├── SKILL.md # Workflow the agent follows
├── references/ # Loaded on demand
│ ├── 01-apple-process.md
│ ├── 02-case-studies.md
│ ├── 03-rules-and-typography.md
│ └── 04-drawing-and-tests.md # Vectorized drawing, scale and context tests, traps
├── examples/ # Worked examples, also the README images
│ ├── cat.md # Critique and refresh of the cat, with the commands' numbers
│ └── sketch-to-concepts.md # Three different concepts from one sketch
└── scripts/ # Node 24, no dependencies
├── scale-report.ts # On-screen size of every gap, shape and stroke at 256, 48 and 16 px
└── render-png.ts # PNGs at 256, 48 and 16 px and a sheet to look at before showing
Disclaimer
Unofficial project, not affiliated with or endorsed by Rob Janoff. The references are original summaries of the book with short quotations; no text or files from the book are included.
License
Reviews (0)
Sign in to leave a review.
Leave a reviewNo results found