Logo Design Skill
Turn a brand brief into tested SVG logo concepts, then build a brand kit after you choose a direction.
- Source repo
- kaankiziltug/logo-design-skill
- Stars
- ★ 7.5k
- Last updated
- 8d ago
- License
- MIT
- Primary language
- HTML
- FA score
- 66/100 · Some gaps
At a glance
- How it runs
- Works with
- Universal · cross-platformCodex · Claude Code · Claude.ai
- Cost
- Free, no paid service needed
- Setup effort
- Low · running in minutes
- You'll need
- Typical use
- An independent designer needs to turn a new brand brief into logo directions and prepare a delivery kit after the client picks a concept.
- Not a fit if
- Teams expecting a complete brand kit without choosing a concept
- Users who only need general image generation, not SVG and identity work
- Maintenance
- Commits in the last 30 dayslast commit 2026-11-13latest release v1.4.5 (2026-10-20)+77.6 ★/day over 7 days
- Source review
- 66/100 · Some gaps
What does this agent do, and when should you use it?
This logo-design skill guides Claude and other agents that support Agent Skills through a process from discovery to production-ready SVGs. It includes SKILL.md, on-demand design references, Python scripts, and a reference library of more than 1,400 real-world SVG logos. The workflow researches category examples, develops 8–12 concept directions, builds three, and tests them through audits and size and colour treatments. It stops at a concept checkpoint; colour, lockups, presentation boards, icons, and guidelines follow only after the user chooses a direction and asks for the kit. Install the skill folder in a supported agent’s skills directory; the scripts use Python’s standard library, while PNG/ICO export needs an available renderer.
The skill uses a supplied brief or stated assumptions for discovery and word mapping, then searches reference marks by technique or industry with scripts/search_library.py. It guides the agent to propose 8–12 concepts and build three as SVGs, then uses scripts/svg_audit.py to check structure, colours, complexity, angles, tiny details, and centring. scripts/concept_sheet.py assembles the concepts; scripts/preview_sheet.py creates tests for size, one-colour and reversed treatments, contexts, and competitor shelves. The checkpoint presents the concepts and a recommendation, then waits for the user to choose a direction. After approval, it can produce brand colours, lockups, presentation boards, icons, and guidelines, with scripts/render_png.py and scripts/export_variants.py available for rendering and export.
- An independent designer needs to turn a new brand brief into logo directions and prepare a delivery kit after the client picks a concept.
- A small product team wants to check whether an SVG mark remains clear at 16 px and in one-colour or reversed treatments.
- A brand consultant wants to search reference marks by industry and technique before designing, to study conventions and avoid look-alikes.
- A team with an existing mark needs an SVG audit, a test sheet, or exported favicon and app-icon variants.
Key setup facts
Pulled from the install and usage notes below. The project README has the current values.
- Config files
~/.codex/skills/logo-design~/.gemini/skills/logo-design- Runtime
- Python 3.8+
- Install notes cover
- macOS
How do you install or deploy this agent?
For Claude Code, install through the plugin marketplace:
/plugin marketplace add kaankiziltug/logo-design-skill
/plugin install logo-design@logo-design-skillAlternatively, clone the repository and copy the skill folder:
git clone https://github.com/kaankiziltug/logo-design-skill.git
cp -r logo-design-skill/skills/logo-design ~/.codex/skills/logo-designFor Gemini CLI, the documented personal directory is ~/.gemini/skills/logo-design; for other Agent Skills-compatible agents, follow their skills-directory documentation. Claude.ai and Claude Desktop can use logo-design.zip from Releases or a package created with python3 tools/package_skill.py; logo-design-lite.zip omits the SVG files. Scripts require Python 3.8+. PNG/ICO export requires one available renderer: cairosvg, rsvg-convert, Inkscape, a Chromium-based browser, or macOS Quick Look.
How do you use this agent?
Start a new session after installation and ask for a logo, wordmark, monogram, app icon, favicon, rebrand, or critique. For example:
Design a logo for Harbor, a savings app for first-time savers. It should feel calm and safe.You can also run scripts from skills/logo-design:
python3 scripts/search_library.py --technique negative-space --exemplary
python3 scripts/svg_audit.py my-logo.svg
python3 scripts/concept_sheet.py a.svg b.svg c.svg --names "A" "B" "C" --recommend 1 -o concepts.png
python3 scripts/preview_sheet.py my-logo.svg --refs-industry developer-tools -o preview.html
python3 scripts/render_png.py my-logo.svg --size 512 -o my-logo.png
python3 scripts/export_variants.py my-logo.svg --mono "#0F7C80" --icon-bg "#0F7C80" --web-iconsThe skill presents concepts and pauses at the checkpoint. It continues with the full brand deliverables only after you choose a direction and request the kit.
What are this agent's strengths and limitations?
- A defined user-choice checkpoint lets the client select a concept before the full kit is produced.
- The library contains 1,400+ classified SVG logos searchable across mark types, techniques, geometry, and industry.
- Standard-library Python scripts cover SVG audits, concept and test sheets, presentation boards, and icon-variant exports.
- The testing workflow includes 16 px, one-colour, reversed, contextual, and competitor-shelf checks.
- The workflow works best with an agent that can view images, since it renders and checks its drafts.
- PNG/ICO export requires an additional renderer: cairosvg, rsvg-convert, Inkscape, a Chromium-based browser, or macOS Quick Look.
- Logos in the reference library are their owners’ trademarks and are excluded from the MIT license; users must follow their separate terms.
- Full-kit production waits for the user to choose a concept at the checkpoint, so it does not suit unattended end-to-end runs.
How does this agent compare with similar options?
Key facts side by side with the most closely related agents.
| Agent | Source review | Form / cost | Stars | Updated | Language | Full support on |
|---|---|---|---|---|---|---|
| Logo Design Skill This agent | 66 · Some gaps | Agent plugin / skillFree | ★ 7.5k | 8d ago | HTML | Codex · Claude Code · Claude.ai |
| Designer Skills Pack | 57 · Major gaps | Agent plugin / skillFree | ★ 3.5k | 2mo ago | Markdown | Claude Code |
| Taste-Skill: AI Frontend Aesthetic Upgrade | 72 · Some gaps | CLIFree + model costs | ★ 120k | 13d ago | JavaScript | ChatGPT · Codex · Claude Code |
| Kubai Icons | 48 · Major gaps | Agent plugin / skillFree + model costs | ★ 1.1k | 1d ago | JavaScript | Codex · Claude Code |
How does FollowAgents rate this agent?
Why each dimension lost points
The tools are described as dependency-free Python, and the test workflow is limited to contents: read, supporting a modest least-privilege and dependency-security assessment (least_privilege 2, dependency_security 2). The process pauses after presenting concepts and does not build the full kit without user selection (user_confirmation 3). The skill explains local scripts and reference-library use, but does not fully explain data processing and flows beyond generated artifacts (data_flow_transparency 2); dedicated handling guidance for sensitive inputs is not shown (sensitive_data_handling 1). The docs mention that PNG rendering may use Chrome, while the traffic archive workflow uses a token and pushes to a private data repository. These external effects have only partial explanation, and general rollback guidance is not shown (external_effects 1, rollback 1). An author and MIT license are identified, but the supplied material does not adequately establish attribution rules for reference marks (source_attribution 1).
README, SKILL, and plugin metadata are broadly consistent about the product. The test workflow shows a multi-platform Python matrix and unit tests, but this static review did not execute them, and the workflow cannot establish correctness of design outputs (self_consistency 2). Python tools are described as having no extra dependencies, and Windows Python launcher differences are documented; rendering may still require Chrome, and dependency availability is not fully covered (dependency_availability 2). The materials describe audit checks but do not sufficiently show diagnostic messages or recovery paths (failure_messages 1).
The skill covers new design, critique, redesign, identity systems, and assets, with multiple mark types and a broad range of industry examples (audience_and_scenarios 3). It defines a boundary between concept and full-kit stages and allows users to request skipping the checkpoint, though capability limits are not comprehensively stated (capability_boundaries 2). Its trigger description explicitly covers logo, favicon, branding, and related requests, while also potentially encompassing broad branding work (trigger_precision 2). The README gives installation paths for Claude, Gemini CLI, Codex CLI, and others, and notes that an image-capable model works best; compatibility across environments is still described mostly in general terms (environment_fit 2).
The README has contents, workflow, installation, examples, and tool descriptions, while the skill is organized by mode, phase, and tool; the supplied README excerpt is truncated, limiting assessment of full navigation (information_architecture 2). Installation instructions cover plugin, manual, and multiple Agent setups in concrete detail (install_notes 3). The skill name is consistent across README, plugin metadata, and front matter, but the only supplied version record is 1.4.5 in plugin manifests (naming_stability 2, versioning_changelog 1). Examples are numerous and span industries, but no FAQ or explicit known-limitations list is shown; Windows launcher and image-model notes do not amount to comprehensive limitation guidance (examples_and_faq 2, known_limitations 1). The repository includes the full MIT text (license 3). An author field is present, but no ongoing maintenance responsibility or update commitment is described (maintenance_responsibility 1).
The stated outputs include SVGs, concept sheets, test sheets, presentation boards, and icon sets; the README provides a reviewable process and examples, though static evidence cannot establish actual output quality (output_usability 3). The classified reference library, SVG audit, size/background checks, and user decision checkpoint provide concrete value beyond a prompt alone (marginal_value 3). Dependency-free Python and several output tools support a favorable ordinary-use cost profile, but rendering may require Chrome and installation, run, and maintenance costs are not quantified (cost_benefit 2).
The README gives specific feature and process claims, the SKILL lists scripts and behavior, and workflow/test files offer partial implementation evidence; however, the packet is only a sample and key source files are not shown (claim_traceability 2). README, skill instructions, plugin manifests, and CI/test materials corroborate product identity and basic capabilities, while the showcased results are primarily repository-reported (cross_source_corroboration 2). The docs distinguish the concept checkpoint, recommendations, and example choices, and mention some risks; claims such as the 1,400+ library and professional results rely mainly on repository assertions and cannot be independently confirmed from this static sample (fact_inference_separation 2).
- The example images, 1,400+ logo-library count, and claimed design outcomes are not independently verified in the supplied sample; check source and licensing before using reference marks.
- For PNG rendering or traffic archiving, account for the Chrome requirement and the traffic workflow’s use of a token and writes to a private repository.
Recent releases
Latest entries from the project's GitHub Releases. Notes are quoted from upstream as written, not rewritten or tested by FollowAgents.
-
Release notes excerpt
Bug fixes for exports and presentation boards. - Exported SVG titles are escaped. A --title with an ampersand, such as "Marlow & Finch", produced SVG files that wouldn't open, and markup in a title could turn into SVG elements. Titles are now escaped in every exported variant. Thanks @andromia3 (#6)! - Mockup handles are clean. Presentation boards turned brand names with punctuation into handles like alex@marlow&finch.com. Handles, e-mail addresses and domains now keep only letters and digits, with accents folded ("Café" → "cafe"). …View on GitHub ↗ -
Release notes excerpt
Reliable rendering on Windows, automated tests, and install guides for other agents. - Windows fix: web-icon export (export_variants.py --web-icons) could stop without writing favicon.ico and the app icons, because Windows keeps Chrome's helper processes and files locked for a moment after a screenshot. render_png.py now stops the whole browser process tree, cleans up temporary folders without failing, and retries a failed browser render once. The export now names the icon that failed instead of saying no renderer is available. …View on GitHub ↗ -
Release notes excerpt
Windows fix. - Scripts no longer crash on Windows consoles (cp1252) when printing report symbols such as ▲ and →. Thanks @RossWarner21 (#1)! - SKILL.md: on Windows, use python or py -3 instead of python3 (often the Microsoft Store stub). - render_png.py no longer writes __pycache__ into the skill folder. Update in Claude Code: /plugin marketplace update logo-design-skill Claude.ai / Desktop: download logo-design.zip (or logo-design-lite.zip) and re-upload it under Settings → Capabilities → Skills.View on GitHub ↗
FAQ
Which agents can use it?
Do I need an API key or paid service?
Will it automatically produce a complete brand kit?
What do the scripts require?
Is Logo Design Skill open source?
Is Logo Design Skill safe to use?
Is Logo Design Skill still maintained?
Who makes Logo Design Skill?
Thanks — we use this to improve the page.