icons
Skill2 starsChanged 3 months agoMITView on GitHub
Hybrid icon system. First tries to source every needed icon from ONE library (Lucide / Phosphor / Heroicons) unified to the DESIGN-LOCK style spec. Falls back to AI-generated SVG only when the library lacks a themed concept, with strict geometry validation to prevent broken paths and hallucinated artifacts.
10 other projects use a version of this.
Adopted by
Most recently taken up by
- stritar/skillstheir own version, first seen Nov 16, 2026
- microsoft/coreai-design-systemtheir own version, first seen Nov 15, 2026
- laguagu/claude-code-nextjs-skillstheir own version, first seen Nov 14, 2026
- JetBrains/intellij-communitytheir own version, first seen Nov 13, 2026
- raphaelsalaja/skillstheir own version, first seen Nov 8, 2026
- github/replicatheir own version, first seen Nov 1, 2026
Other versions10 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 10
- JetBrains/intellij-community★ 21ksame name
- DetachHead/rebased★ 6.5ksame name
- github/replica★ 275same name
- laguagu/claude-code-nextjs-skills★ 72same name
- raphaelsalaja/skills★ 67same name
- microsoft/coreai-design-system★ 2same name
- DegreeSign/skills★ 1same name
- alexjmiller5/agent-config-public★ 0same name
- AsimSayed/claude-skills★ 0same name
- stritar/skills★ 0same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository15 other files it gives its agents.
- CLAUDE.mdCLAUDE.md
- animationSkill
- auditSkill
- blocksSkill
- briefSkill
- buildSkill
- copy-analyzeSkill
- deploySkill
- design-lockSkill
- inboxSkill
- mediaSkill
- mobileSkill
- refsSkill
- tokensSkill
- typographySkill
The file
Show its rules and text
We split this skill into the rules it gives an agent
14 rules. 0 are common, 9 are distinctive, and 5 are very specific to this project.
0
Common
Most of the 9 edited versions we read keep these.
9
Distinctive
This author's own choices. Worth a look.
5
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Very specific: line 12
After /media — media assets are in place, now icons are the last visual asset.
- Very specific: line 13
Before /build — build references the icon set from site/media/icons/.
- Distinctive: line 47
Read DESIGN-LOCK and derive:
- Distinctive: line 112
Validate: Run the SVG through these gates:
- Very specific: line 116
Paths closed: Every <path d="…"> either ends with Z/z or returns to its M start within 0.5 units.
- Distinctive: line 120
Symmetry (conditional): If concept implies symmetry (heart, star, target, shield, cross) — measure axis symmetry within 2% tolerance.
- Distinctive: line 133
All icons — library-sourced and AI-generated — go through the same normalization:
- Distinctive: line 195
No decorative icons: Every icon must earn its place tied to a concept from the copy.
- Very specific: line 196
currentColoronly: Never hardcode fill/stroke in SVG — inherit from CSS so tokens.css controls color. - Distinctive: line 201
Do not generate icons before locking the library.
- Distinctive: line 202
Do not accept AI output that failed a validation gate — retry or fall back to library.
- Very specific: line 203
Do not embed inline <style> in SVGs — all styling via CSS classes.
- Distinctive: line 204
Do not use icon fonts (FontAwesome, Material Icons) — SVG only, unified pipeline.
- Distinctive: line 205
Do not add icons for concepts the user did not confirm in Step 1.
“Common” is measured: kept by at least 6 in 10 of the 9 edited versions of this file we could read. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
/icons
Icons are a common failure point: mixed libraries, mismatched stroke weights, decorative-only icons, hallucinated SVG shapes. This skill enforces one library, one style, verified geometry.
When to use
- After
/media— media assets are in place, now icons are the last visual asset. - Before
/build— build references the icon set fromsite/media/icons/. - Rerun anytime a new concept is added to the copy.
Requirements
DESIGN-LOCK.mdLOCKED (need §2 type, §3 color, §5 motion, §7 imagery for style spec).brief/copy-analysis.mdexists (concepts are extracted from here).inbox/texts/populated (backup source for concept extraction).
Process
Step 1. Extract concepts from context
Read:
brief/01-business.md— niche, product, services.brief/copy-analysis.md— recommended section structure and key claims.brief/03-goals.md— CTA labels and offer names.inbox/texts/*— surface-level scan for repeating nouns.
Extract 5-15 icon concepts. Each concept is a short label + intent:
- consulting → conversation, expertise
- roadmap → path, planning
- reporting → chart, transparency
- fast-turnaround → clock, speed
- privacy → lock, safety
- multi-region → globe, coveragePresent the extracted list to the user. Ask: "add, remove, or rename any concepts before we map to icons?"
Step 2. Compute style spec from DESIGN-LOCK
Read DESIGN-LOCK and derive:
Style spec
stroke-width: <derived from type §2 body-weight and radius §4>
thin (1px) if radius = 0
regular (1.5px) if radius ≤ 8
bold (2px) if radius ≥ 12 or pill
corner-style: sharp if radius = 0 else rounded
fill-style: outline (default) | filled (only if brand uses heavy typography)
| duotone (only if palette §3 is duotone/tritone)
size-baseline: 24×24 viewBox
color-strategy: currentColor (inherit from text via CSS)
geometric-complexity: minimal (≤ 2 paths) | detailed (up to 4 paths)Show the derived spec. Ask user to confirm or override.
Step 3. Library selection
Present three candidate libraries, ranked by fit against the style spec:
| Library | Best for | Stroke default | License |
| Lucide | tech, SaaS, minimal | 2px | ISC |
| Phosphor | editorial, brand, versatile (6 weights) | 1.5px thin | MIT |
| Heroicons | minimal, product | 1.5px outline / 24px solid | MIT |
Auto-recommend one based on style spec match. User picks. This choice is written to DESIGN-LOCK §7 (Icons field) and becomes locked for the project — no mixing.
Step 4. Concept-to-icon mapping
For each concept, find the closest library match. Present in a table:
concept → library icon confidence
consulting → phosphor/chats-teardrop high
roadmap → phosphor/map-trifold high
reporting → phosphor/chart-line-up high
fast-turnaround → phosphor/clock medium
privacy → phosphor/lock-key high
multi-region → phosphor/globe-hemisphere-west medium
custom-integration → (none) MISSING → AI fallbackConfidence levels: high (exact match), medium (adjacent concept), low (loose semantic tie), MISSING (no plausible match — flag for AI fallback).
User reviews. For each row user can: accept, pick a different library icon (show 3 alternatives), or move to MISSING to force AI generation.
Step 5. AI fallback for MISSING concepts only
Never generate an icon if a library option exists. AI is last resort.
For each MISSING concept:
- Build a strict prompt. Include:
- Concept meaning (one sentence).
- Style spec (stroke-width, corner-style, fill-style).
- Size:
viewBox="0 0 24 24". - Color:
stroke="currentColor" fill="none"(orfill="currentColor"for solid). - Constraints: max 2 paths, no gradients, no filters, no text elements, all paths closed, symmetric where the concept implies it.
- Reference: "match visual weight of Phosphor Regular icons" (or whichever library was picked).
- Generate. Produce a candidate SVG.
- Validate. Run the SVG through these gates:
- Parseable. Valid XML, no unclosed tags. If fail → retry (max 3).
- viewBox. Exactly
0 0 24 24. If fail → rewrite. - Path count.
<= 2<path>elements (or up to 4 ifgeometric-complexity=detailed). If fail → retry with tighter prompt. - Paths closed. Every
<path d="…">either ends withZ/zor returns to itsMstart within 0.5 units. If fail → retry. - Stroke width match. All
stroke-widthattributes equal library baseline. If mismatch → normalize. - No prohibited elements. No
<text>,<image>,<filter>,<mask>,<foreignObject>,<defs>with gradients. If fail → retry. - No dangling points. Detect degenerate
M x,y M x,ysequences with no intervening geometry. If fail → retry. - Symmetry (conditional). If concept implies symmetry (heart, star, target, shield, cross) — measure axis symmetry within 2% tolerance. If fail → retry.
- Retry policy. Up to 3 attempts. If all fail, escalate to user:
⚠️ Could not generate a clean icon for "custom-integration" after 3 attempts. Options: A. Pick a library alternative (show 3 loose matches from Phosphor) B. Refine the concept ("integration" → "handshake"? "puzzle"?) C. Skip — this concept ships without an icon
Step 6. Unification pass
All icons — library-sourced and AI-generated — go through the same normalization:
- SVGO with a locked config:
{ "plugins": [ "preset-default", { "name": "removeViewBox", "active": false }, { "name": "removeDimensions", "active": true } ] } - Inject
stroke="currentColor"andfill="none"(or the fill counterpart for solid) — remove hardcoded colors. - Enforce
stroke-widthto the library baseline. - Add
class="icon icon-<concept>"to the root<svg>for CSS targeting. - Add
aria-hidden="true"by default. If the icon is standalone (no accompanying text), it must instead haverole="img"and<title>— flag for the user to decide.
Step 7. Output
Write to site/media/icons/:
site/media/icons/
consulting.svg
roadmap.svg
reporting.svg
fast-turnaround.svg
privacy.svg
multi-region.svg
custom-integration.svg ← AI-generated, marked in README
README.md ← concept → file map + source (library or AI)
_snippets.html ← ready-to-paste <svg> inline snippets for /buildsite/media/icons/README.md structure:
# Icons
Library: **Phosphor Regular** (locked in DESIGN-LOCK §7).
Total: 7 icons (6 library, 1 AI-generated).
| concept | file | source | notes |
|---------|------|--------|-------|
| consulting | consulting.svg | phosphor/chats-teardrop | — |
| ... |
| custom-integration | custom-integration.svg | AI-generated | validated: 2 paths, symmetric |Append to references/approved-manifest.md — every icon that ships is a required element in the corresponding section:
[ ] icon:consulting → services section
[ ] icon:privacy → footer/trust/build confirms each icon appears in the generated HTML.
Rules of behavior
- One library, no mixing. Once picked in Step 3, no other library icons enter the project.
- AI only when necessary. Library-first, always. AI fallback is the last resort, gated by validation.
- No decorative icons. Every icon must earn its place tied to a concept from the copy. Decoration = anti-slop rule violation.
currentColoronly. Never hardcode fill/stroke in SVG — inherit from CSS so tokens.css controls color.- Validate before shipping. An AI SVG without geometry validation is a broken-icon bug waiting to happen.
What NOT to do
- Do not generate icons before locking the library.
- Do not accept AI output that failed a validation gate — retry or fall back to library.
- Do not embed inline
<style>in SVGs — all styling via CSS classes. - Do not use icon fonts (FontAwesome, Material Icons) — SVG only, unified pipeline.
- Do not add icons for concepts the user did not confirm in Step 1.
Next
/build— readssite/media/icons/_snippets.htmland drops inline SVG into sections. Checks approved-manifest coverage./audit— confirms only one icon library was used, all icons havearia-hiddenorrole="img", no icon fonts loaded.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

