# logo-design-skill: structured logo pipeline for Claude and AI agents > logo-design-skill is a Claude and AI agent plugin that guides brief-to-SVG logo work through a staged process, with a mandatory approval checkpoint and a 1,400+ SVG reference library for category research. The Python test tools are dependency-free, but the full package is large enough to require a lite version for Claude.ai uploads. **kaankiziltug/logo-design-skill** — A comprehensive logo-design skill for Claude, Gemini CLI, Codex and other AI agents: principles, process, SVG craft, testing tools and a 1,400+ logo reference library. - Repository: https://github.com/kaankiziltug/logo-design-skill - Stars: 2,743 · Forks: 197 - Language: HTML - License: MIT - Published: 2026-10-11 · Updated: 2026-10-11 · Language: en - Canonical page: https://hysenlabs.com/projects/kaankiziltug-logo-design-skill ## Brief questions to production SVG: what the skill pipeline covers logo-design-skill describes itself as turning Claude, or any agent that supports Agent Skills, into a disciplined identity designer. Scope runs from the first brief through discovery, word mapping, mark type selection, concepting, geometric construction and optical corrections that include overshoot, bone effect and irradiation. Colour, typography, lockups, testing, presentation, delivery, identity systems and redesigns are all in scope. Administrators and designers who want structured logo work from an AI agent, rather than a free-form prompt, are its primary audience. Without a skill like this, an agent producing SVG logos has no built-in knowledge of category conventions, optical corrections or the concept approval step. Repository topics name the agents it targets: claude, gemini-cli and codex alongside agent-skills and logo-design. ## A mandatory concept checkpoint before the full kit runs Stages run in sequence: brief, research, concepts, test and refine, then a checkpoint, then the kit. Generating 8 to 12 concept one-liners comes first, followed by building 3 of them as SVG files, running them through audit and test tools, and stopping at the checkpoint. Only after a direction is picked does the full kit run, because producing a complete brand kit before a concept is approved uses most of the session's token budget for work that may be discarded. Among the 28 fictional brand examples in the repository, 4 had the person pick a direction different from the skill's recommendation, after which the skill refined that choice before colouring it. That scenario is exactly what the checkpoint exists to handle. ## Installing on Claude Code, Gemini CLI, Codex CLI and Claude.ai For Claude Code, the recommended path uses the plugin marketplace: ```bash /plugin marketplace add kaankiziltug/logo-design-skill /plugin install logo-design@logo-design-skill ``` A manual install clones the repository and copies the skills folder. Personal installs go to `~/.claude/skills/logo-design` for Claude Code, `~/.gemini/skills/logo-design` for Gemini CLI, and `~/.codex/skills/logo-design` for Codex CLI: ```bash git clone https://github.com/kaankiziltug/logo-design-skill.git cp -r logo-design-skill/skills/logo-design ~/.claude/skills/logo-design ``` For Claude.ai or Claude Desktop, both `logo-design.zip` and `logo-design-lite.zip` are on the project's Releases page. Upload under Settings → Capabilities → Skills. Lite omits the SVG reference files and targets uploads that hit a size limit. Starting a new agent session after any install method is enough: the agent picks the skill up from its description without additional configuration. ## Dependency-free Python tools for SVG audit, test sheets and icon export Python scripts in the tools directory are described as dependency-free. Four categories of work are documented. Audit checks an SVG against logo design principles and reports problems. Test sheets generate seven standardized views: 16 px pixel test, one-colour, reversed, squint, mirror, contexts and competitor shelf test. These expose whether a mark reads at small sizes, works without colour, survives reversal and holds its own next to competitors on a shelf. Client presentation boards draw on industry-specific mockups. PNG rendering and a complete favicon, app-icon and web-manifest export are also documented outputs. Running `python3 tools/package_skill.py` produces the zip file for uploading to Claude.ai, so a fresh build from source is an alternative to the Releases page download. Plain Python and the same folder structure work across all supported agents without modification. ## Seven classification dimensions for 1,400+ logos, with research as the only permitted use More than 1,400 real-world SVG logos make up the reference library. Each entry is classified across seven dimensions: mark type, technique, geometry, subject, typography, mood and industry. A command-line search interface and a local gallery browser both come with the repository. Purpose is explicitly bounded. Classification data serves category research and look-alike avoidance, identifying what visual territory a given sector already occupies. Copying any logo in the library is outside the stated scope, a boundary the README names directly. This library is what makes the main zip large. Lite strips it out, leaving skill instructions and Python tools but removing the research function. An operator who installs lite to meet a size limit accepts that the category research stage of every brief will run without it. ## Image viewing, the .claude-plugin folder and the Agent Skills format Working best with a model that can view images is a condition the README names, because drafts are rendered to PNG for a self-check before they are shown. Agents running models without image viewing capability can still load the skill but skip that rendered self-check. `.claude-plugin/` exists for Claude Code-specific integration and is inert for all other agents. Cloning the repository and copying the skills folder into Gemini CLI or Codex CLI requires no cleanup: files irrelevant to those agents have no effect. Agent Skills is an open format: a folder containing a `SKILL.md` file with Markdown instructions alongside Python tools. Cursor and GitHub Copilot are listed as supported agents alongside Gemini CLI and Codex CLI. Any unlisted agent that accepts a skills folder in this format can use logo-design-skill. ## When a staged pipeline and a large library are more than the project needs A mandatory mid-process stop makes logo-design-skill unsuitable when a finished kit from a single uninterrupted run is the requirement, and no parameter or documented workaround in the README describes a way to skip the concept review stage. Package size is an operational consideration that directly affects which install path works. Installing the full package on Claude.ai requires either accepting the upload as-is or switching to the lite version; lite removes the visual reference library that the research stage depends on. For simple one-off wordmarks where category research and a seven-view test sheet are unnecessary, a plain SVG prompt produces output without the staged overhead. Agents that do not support the Agent Skills format and a SKILL.md file cannot use logo-design-skill regardless of model capability. Checking format support before install is the first step for any unlisted agent. ## Conclusion logo-design-skill is worth installing if you use Claude, Gemini CLI or Codex CLI and want structured logo work with built-in category research, optical correction guidance and SVG test tooling. A mandatory mid-process stop is part of the design: no full brand kit is generated until a concept direction is approved. Before installing for Claude.ai or Desktop, check the upload size limit and decide whether the lite package is required; if you use lite, the 1,400+ SVG reference library is absent. Confirm your agent model supports image viewing, since the self-check step renders drafts to PNG before showing them. ## FAQ ### What is logo-design-skill? logo-design-skill is a Claude and AI agent plugin that guides brief-to-SVG logo design through a structured process. It covers brief discovery, concept generation, SVG construction, test tooling and brand kit delivery, and includes a reference library of more than 1,400 classified real-world logos. ### How do I install logo-design-skill for Claude Code? The recommended path uses the plugin marketplace: /plugin marketplace add kaankiziltug/logo-design-skill, then /plugin install logo-design@logo-design-skill. A manual install clones the repository and copies the skills folder to ~/.claude/skills/logo-design for a personal install. ### Does logo-design-skill work with Gemini CLI? Yes. logo-design-skill uses the open Agent Skills format, which Gemini CLI supports. Clone the repository and copy the skills folder to ~/.gemini/skills/logo-design, then start a new session. ### What does logo-design-lite.zip contain? logo-design-lite.zip includes the skill instructions and Python tools but omits the 1,400+ SVG reference library. It targets uploads that hit a size limit on Claude.ai or Claude Desktop. ### What happens at the checkpoint in logo-design-skill? A greyscale concept overview image and a recommendation are shown, and no further output is generated until a direction is chosen. Full brand kit generation only starts after that approval. ## Sources - [Issues](https://github.com/kaankiziltug/logo-design-skill/issues) - [kaankiziltug/logo-design-skill on GitHub](https://github.com/kaankiziltug/logo-design-skill) - [License: MIT](https://github.com/kaankiziltug/logo-design-skill/blob/main/LICENSE) - [README](https://github.com/kaankiziltug/logo-design-skill/blob/main/README.md) - [Releases](https://github.com/kaankiziltug/logo-design-skill/releases) --- Hysen Labs editorial analysis, written from the project's own repository and release notes. Cite the canonical page: https://hysenlabs.com/projects/kaankiziltug-logo-design-skill