Notre avis
Ce skill identifie les composants source correspondant à des captures d'écran ou descriptions visuelles d'interfaces utilisateur.
Points forts
- Correspondance précise via plusieurs signaux visuels (texte, BEM, Material, layout, couleur, forme des données).
- Indexation automatique des composants avec mise à jour incrémentielle.
- Intégration en amont de /scout pour cibler les recherches.
- Graphe de dépendances entre composants pour comprendre l'architecture.
Limites
- Nécessite que l'index des composants soit maintenu à jour.
- Dépend de la qualité visuelle de la capture d'écran (résolution, éléments cachés).
- Ne fonctionne que pour les projets déjà configurés avec la structure de dossiers attendue.
Utilisez ce skill lorsqu'un développeur a besoin de localiser rapidement le code source d'un élément UI visible dans une capture d'écran.
Ne l'utilisez pas pour des modifications de code sans référence visuelle, ou lorsque la capture d'écran ne montre pas clairement des composants spécifiques.
Analyse de sécurité
SûrThe skill runs local Node.js scripts and uses Bash to generate a component index, then reads JSON and source files. It does not perform network operations, data exfiltration, or destructive actions. All tools are used within the user's own codebase for legitimate UI component discovery.
Aucun point d'attention détecté
Exemples
I have a screenshot of a login button with a Material Design style. Find the component responsible for it.Find the page component that renders the dashboard layout shown in this screenshot.Update the component index after recent edits to shared components.name: find-component description: >- [Frontend] Find source code components from UI screenshots or visual descriptions. Runs BEFORE /scout when screenshot is pasted with UI intent. Triggers on: find component, find this component, where is this code, where is this page, which component, locate component, find the source, what component is this, refresh component index, update component index, sync component index. Maps visual signals to source files with confidence scoring and Mermaid component graph. allowed-tools: Read, Grep, Glob, Bash, Task
Anti-Hallucination Requirement
Read: .claude/skills/shared/anti-hallucination-protocol.md
Read: references/confidence-scoring-guide.md
Every claimed match MUST have grep-verified evidence. HALT if confidence <85% — show candidates, never fabricate.
Workflow
Step 1: Load Component Index
# Full scan (rebuild entire index):
node .claude/skills/find-component/scripts/generate-component-index.cjs
# Incremental (only re-scan git-changed component files):
node .claude/skills/find-component/scripts/generate-component-index.cjs --git-changes
# Then read:
Read: docs/component-index.json
When to use which mode:
--git-changes— fast update after editing components (merges into existing index)- No flag — full rebuild when index missing, stale, or after major refactors
- Fallback: live grep if index unavailable
Step 2: Extract Visual Signals from Screenshot
Read: references/visual-signal-extraction.md
Read the screenshot. Extract 6 signal types (text, BEM, Material, layout, color, data shape). List each signal with what you ACTUALLY see — never infer.
Step 3: Bottom-Up Component Search
Read: references/component-matching-protocol.md
- Reusable components first — match signals against
platform-core/andlibs/components - Domain components — match against
shared/components/ - Page component — find parent that composes matched children
- If no reusable match → search page templates directly
Step 4: Score Confidence
Read: references/confidence-scoring-guide.md
- ≥85% (≥2 grep-verified signals) → report as confirmed match
- <85% → show top candidates with per-signal breakdown, ask user
Step 5: Build Component Relationship Graph
Output Mermaid diagram:
graph TD
A[PageComponent] --> B[ChildComponent1]
A --> C[ChildComponent2]
B --> D[SharedComponent]
Verify each edge by reading actual template HTML. Never infer from imports alone.
Step 6: Report
Output: matched components, file paths (TS/HTML/SCSS), confidence %, Mermaid graph, and any unresolved ambiguities.
Integration — Pre-Scout Activation (IMPORTANT)
When a screenshot is pasted with UI-related intent (fix, update, find, debug), this skill
runs BEFORE /scout in any workflow. It narrows the search scope so scout is targeted.
Auto-activation priority: /find-component → /scout → remaining workflow steps
Applies to: bugfix, feature, refactor, fix-ui, verification workflows.
Output (matched file paths + component graph) feeds directly into scout/investigate steps.
Expert Next.js App Router
Developpement
Un skill qui transforme Claude en expert Next.js App Router.
Générateur de README
Developpement
Crée des README.md professionnels et complets pour vos projets.
Rédacteur de Documentation API
Developpement
Génère de la documentation API complète au format OpenAPI/Swagger.