logo-designer
Design and iterate on logos using SVG. Use this skill when the user asks to "create a logo", "design a logo", "make me a logo", "iterate on this logo", "logo for my project", or discusses logo design, branding icons, or wordmarks.
By neonwatty · 721 installs
npx skills add neonwatty/logo-designer-skill --skill logo-designer
Source repository · Upstream listing
Logo Designer
Design and iterate on logos using SVG. Generates side by side previews and exports to PNG at standard sizes.
Phase 1: Interview
Before generating anything, gather context and ask the user what they need.
Step 1: Gather context automatically
If the user points to a repo, URL, or existing project:
Read the README, package.json, CSS/config files, and any existing branding
Extract: project name, purpose, tech stack, color palette, design language, fonts
Summarize what you found before asking questions — this avoids asking things you already know
If the user just says "design a logo" with no project context, skip to Step 2.
Step 2: Ask structured questions
Use the AskUserQuestion tool to ask these questions. Batch related questions together (up to 4 per call) and skip any question already answered by the context gathered in Step 1 or by the user's initial message.
Question 1 — Format:
Question 2 — Style direction:
Question 3 — Color preferences:
Question 4 — Output size (only if the user mentioned a specific platform):
Adapting to context
User points to a repo: Gather context first, then ask only format + style (colors are likely known).
User says "design a logo for X": Ask format, style, and colors together.
User gives detailed description: Skip everything already covered, ask only what's missing.
User says "just make something": Use sensible defaults (icon only, minimal, surprise me) and go straight to Phase 2.
Move to Phase 2 once you have enough to generate distinct concepts.
SVG Conventions
When generating SVG logos, follow these rules:
viewBox sizing — Always use viewBox="0 0 W H" without fixed width / height attributes. Use 512x512 for icons, 1024x512 for wordmarks/combination marks.
Self contained — No external fonts, images, or <use references to other files. Everything inline.
Text handling — Use widely available system fonts (Arial, Helvetica, Georgia, etc.) or convert text to <path elements. When using system fonts, always include a generic fallback (e.g., font family="Helvetica, Arial, sans serif" ).
Meaningful groups — Wrap logical sections in <g elements with descriptive IDs: id="icon" , id="wordmark" , id="tagline" . This makes iteration easier when the user says "make the icon bigger" or "change the wordmark color".
Flat fills by default — Use solid fill colors. Only use gradients ( <linearGradient , <radialGradient ) when the user requests them or the style clearly calls for it.
Small size legibility — Logos must work at 16 32px (favicons). Prefer solid fills over thin strokes, avoid fine details that disappear at small sizes, and use stroke width of 6+ for any outlines that need to remain visible. Test this mentally: if a detail won't survive being 32px wide, simplify it.
Clean markup — No unnecessary transforms, no empty groups, no default namespace clutter. Keep the SVG readable.
Phase 2: Explore
Generate 3 5 distinct SVG logo concepts. Each concept should take a meaningfully different creative direction — vary the icon metaphor, typography style, layout, or overall aesthetic. Do not generate minor variations of the same idea.
Parallel generation
Use the Task tool to generate all concepts in parallel. This is significantly faster than writing them sequentially.
1. Create the logos/concepts/ directory first
2. Dispatch one Task agent per concept, all in the same message so they run concurrently. Each agent should:
Receive the full design brief (format, style, colors, viewBox, SVG conventions)
Be assigned a specific creative direction (e.g., "geometric letterform", "abstract symbol", "mascot based")
Write its SVG to a specific file path (e.g., logos/concepts/concept 1.svg )
Use subagent type: "general purpose" and mode: "bypassPermissions"
3. After all agents complete, generate logos/preview.html and present the results
Example dispatch pattern (all in one message):
Each agent prompt must include: the full SVG conventions from this skill, the target file path, the specific creative direction, and all relevant context (project name, colors, style preferences). Agents do not share context — give each one everything it needs.
File output
After all parallel agents complete:
1. Generate logos/preview.html using the preview template below
2. Tell the user to open logos/preview.html in their browser
3. Briefly describe each concept (1 sentence each) so the user can match descriptions to visuals
4. Ask: "Which direction do you want to explore? Pick a number, or describe what you like/dislike across them."
Preview HTML Template
When generating logos/preview.html , use this template. Replace {{CARDS}} with one card per SVG file. Set {{PHASE}} to "Concepts" during explore or "Iterations" during refine.
Favicon size check strip
During Phase 3 (Refine), add a "Favicon Size Check" section below the iteration grid. This renders each iteration at 64px, 32px, and 16px so the user can spot legibility issues early. Use this HTML pattern:
For icon only logos, {{FAVICON PATH}} is the iteration path. For combination
marks, create a standalone square SVG from the meaningful icon group and use
its path for {{FAVICON PATH}} . Never squeeze the full horizontal wordmark into
the square favicon cells. If details disappear at 32px, suggest simplifying
(remove fine details, thicken strokes, drop decorative elements).
Each {{CARDS}} entry is:
Where {{PATH}} is the relative path from logos/ (e.g., concepts/concept 1.svg or iterations/iteration 3.svg ) and {{LABEL}} is the filename without extension (e.g., "concept 1" or "iteration 3").
During explore , show all concepts. During refine , show all iterations (most recent first).
Phase 3: Refine
Once the user picks a concept direction, iterate on it.
Single vs. batch iterations
Single iteration — When the user gives specific feedback ("make the icon bigger", "change the blue to green"), apply the change directly and write the next iteration SVG yourself.
Batch variations — When exploring multiple directions at once ("try different color palettes", "show me 5 variations of the eye shape", "experiment with bar count"), use the Task tool to generate variations in parallel, just like Phase 2:
1. Dispatch one Task agent per variation, all in the same message
2. Each agent receives: the base SVG content (copy the full SVG inline in the prompt), the specific variation to apply, the target file path, and the full SVG conventions
3. After all agents complete, regenerate logos/preview.html and present the results
Example batch dispatch:
Use subagent type: "general purpose" and mode: "bypassPermissions" for each agent. Always include the full base SVG content in each agent's prompt — agents do not share context.
File output
1. Copy the chosen concept as the starting point — save the first refinement as logos/iterations/iteration 1.svg
2. Apply the user's feedback and save each new version with an incrementing number
3. Regenerate logos/preview.html after each iteration, showing all iterations (most recent first) so the user can compare
4. Tell the user to refresh their browser after each iteration
5. After each iteration, briefly describe what changed and ask for next feedback
Iteration tips
If the user says "go back to iteration N", use that as the new base
If the user wants to compare specific iterations, mention which filenames to look at in the preview
Keep SVG structure consistent across iterations (same group IDs) so the user can track what changed
Use parallel agents for batch exploration (3+ variations), sequential writes for single tweaks
Check small size legibility — After generating iterations, include the favicon size check strip in the preview. If thin strokes vanish at 32px, proactively suggest thickening them. If fine details (clocks, sparkles, thin icons) become unreadable, suggest removing or simplifying them. This saves iteration cycles.
When the user is satisfied, move to Phase 4
Optional Lineage review (explicit opt in only)
Standalone SVG files and logos/preview.html are always the default. Do not look for,
start, or connect to Lineage merely because it may be installed or a runtime descriptor
exists. Use Lineage only after the user explicitly asks for canvas review and provides
the Lineage checkout or adapter command.
For an explicit review, keep the handoff one way and public boundary only. Run the
Lineage adapter with an explicit artifact, selector, and target, then pipe its single
versioned JSON receipt to the bundled stdin only handoff:
The handoff never starts or locates Lineage and accepts no token, API origin, artifact
argument, or connection context. It consumes only the adapter receipt on stdin. On an
accepted receipt, it atomically creates the next collision safe
logos/iterations/iteration N.svg , rereads the published bytes, verifies them, and
prints a metadata only receipt containing iterationPath , bytes , and sha256 .
File data and supported directory metadata are synchronized before that continuation
receipt is emitted. Pre transaction invalid or unavailable adapter receipts contain no
fabricated transaction, source path, or revision identity and remain terminally
consumable by the same handoff.
Continue refinement only from that exact iterationPath , then regenerate
logos/preview.html so the verified iteration remains visible in the normal workflow.
For reverted, rejected, stale, unavailable, conflict, timeout, or invalid receipts,
follow the printed terminal guidance and do not create or reserve an iteration. The
producer waits through temporary editor disconnections so a reconnected canvas cannot
accept the same proposal after this handoff has stopped listening. If an authoritative
accepted receipt cannot be persisted, exit 27 preserves its transaction identity, byte
count, and hash. Fix the local storage problem and rerun the same adapter command with
that transaction ID and the same artifact; do not create a new transaction. Never
automatically resubmit after timeout or conflict. A new submission
must be an explicit user directed action after checking the current canvas state.
If Lineage reports that its local server was replaced during a provisional acceptance,
do not infer a terminal result and do not continue from browser memory. Inspect the
locked canvas and use its explicit Restore previous document recovery action; only
after that exact transaction is resolved may the user explicitly start another handoff.
Phase 4: Export
When the user says "export", "I'm happy with this", "this is the one", or similar:
1. Identify the final iteration SVG (ask the user to confirm which one if ambiguous)
2. Create the logos/export/ directory
3. Copy the final SVG to logos/export/logo.svg . For a combination mark, also
create a standalone square logos/export/icon.svg from its meaningful
icon group. Preserve the icon's appearance and give it a tight square
viewBox ; do not include the wordmark.
4. Run the bundled export script to generate PNGs. Passing an SVG that is
already at its destination is supported:
For a combination mark, pass the standalone icon as the optional third
argument:
The script produces:
logo 16.png
logo 32.png
logo 48.png
logo 192.png
logo 512.png
logo 1024.png
logo 2048.png
When an icon SVG is provided, the script also preserves icon.svg and produces
the matching icon 16.png through icon 2048.png family. Use the icon
assets for favicons and app icons; use the logo assets where the complete
combination mark belongs.
5. Report the results: list all exported files with their sizes
6. If the export script fails (no conversion tool found), tell the user:
"No SVG to PNG converter found. Install one of: npm inst