02 Wiki Custom Skills Guide - Agente di Design AI Gratuito

Every design team has repetitive tasks — the things you do 50 times a week that collectively consume entire workdays. Custom Skills are Lovart's solution: reusable, AI-powered workflows that automate these tasks with a single click or a simple natural language command.

Think of Custom Skills as macros for design. Just as a developer writes a script to automate deployment, you can build a Skill to automate: "Generate 15 social media variants from this hero image, apply our brand kit, add the campaign hashtag, and export as PNG @2x for every platform." One click, 15 files, zero manual work.

Lovart e' l'agente di design AI con 10M+ creatori. Prova Gratis ->

Lovart is the AI design agent trusted by 10M+ creators. Create logos with AI in minutes →

Lovart is the AI design agent trusted by 10M+ creators. Create logos with AI →

Lovart is the AI design agent trusted by 10M+ creators. Create logos with AI in minutes →

Lovart is the world's first AI design agent — complete brand visual systems from one brief. Try Lovart free →

This guide covers the complete Custom Skills system — from simple parameterized commands to complex multi-step automation pipelines.

What Are Custom Skills?

A Custom Skill is a saved sequence of design operations packaged with AI intelligence. Skills can:

Accept inputs: Text, images, data files, brand references
Apply transformations: Generate designs, modify elements, convert formats
Follow conditional logic: If/then branching based on input characteristics
Chain operations: Execute multiple steps in sequence, with each step's output feeding the next
Export automatically: Save to specific folders, formats, and naming conventions

The Custom Skill Anatomy

Every skill has four components:

Skill: "Social Media Batch Generator" ├── Trigger: How you invoke it (command, button, API, schedule) ├── Input Schema: What information it needs ├── Operation Sequence: What it does, step by step └── Output Configuration: What it produces and where

Creating Your First Skill

Skill Builder Interface

Navigate to Skills > Create Skill to open the Skill Builder. You have two creation modes:

Visual Builder (Recommended for Beginners):
A flowchart-style interface where you drag and connect operation blocks. Each block represents one action — generate, resize, add text, apply filter, export, etc.

Code Builder (Advanced):
Write skills in Lovart's Skill Definition Language (SDL), a YAML-based configuration format with embedded AI prompt instructions.

Example: "Auto Resize for Instagram"

A simple skill that takes any design and resizes it for Instagram's three formats.

Visual Builder Setup:

BLOCK 1: "Input" Type: Image Upload Label: "Upload your design" BLOCK 2: "Resize to Feed Post" Type: Resize Target: 1080x1080 Fit: Cover Position: Center Apply Brand Padding: Yes BLOCK 3: "Resize to Story" Type: Resize Target: 1080x1920 Fit: Cover Position: Center Smart Background Fill: Yes (match brand colors) BLOCK 4: "Export All" Type: Export Format: PNG-24 @2x Naming: {{original_name}}_{{format}}.png Destination: /Social/Instagram/ CONNECTIONS: BLOCK 1 → BLOCK 2 → BLOCK 4 BLOCK 1 → BLOCK 3 → BLOCK 4

SDL Equivalent:

skill: name: "Auto Resize for Instagram" description: "Resize any design for Instagram Feed, Story, and Reels" trigger: type: command phrase: "resize for instagram" inputs: - name: source_image type: image required: true operations: - id: resize_feed action: resize input: source_image params: width: 1080 height: 1080 fit: cover smart_padding: true - id: resize_story action: resize input: source_image params: width: 1080 height: 1920 fit: cover smart_background: brand_colors output: format: png scale_factors: [1, 2] naming: "{{source_name}}_{{operation_id}}" destination: "/Social/Instagram/"

Skill Trigger Methods

Skills can be triggered in five ways:

1. Command Trigger

Invoke skills by typing commands in the Lovart interface:

/resize-instagram /social-batch /export-client-assets

Commands support autocomplete and parameter passing:

/social-batch --platforms=instagram,facebook,linkedin --campaign=summer-launch

2. Button Trigger

Pin frequently used skills as buttons in your workspace toolbar. One click executes the skill with default parameters.

3. Context Menu Trigger

Right-click any design and select from your skills list. The skill receives the clicked design as input automatically.

4. API Trigger

Invoke skills programmatically:

curl -X POST https://api.lovart.ai/v2/skills/skill_abc123/execute \ -H "Authorization: Bearer $LOVART_API_KEY" \ -d '{"inputs": {"source_image": "design_xyz.png", "platforms": ["instagram"]}}'

5. Scheduled Trigger

Set skills to run automatically on a schedule. Available on Business and Enterprise plans.

Schedule: Every Monday at 9:00 AM EST Skill: "Weekly Social Media Batch" Input: CSV from /Marketing/Content Calendar/this-week.csv

Advanced Skill Features

Conditional Logic

Skills can branch based on conditions:

operations: - id: detect_content_type action: ai_analyze input: source_image prompt: "Classify this image as: product_photo, lifestyle, text_graphic, or illustration" - id: product_workflow action: resize input: source_image params: background: white padding: 5% condition: "detect_content_type.result == 'product_photo'" - id: lifestyle_workflow action: resize input: source_image params: background: smart_blur padding: 0% condition: "detect_content_type.result == 'lifestyle'"

AI-Powered Operations

Beyond basic transformations, skills can include AI operations:

Smart Crop: "Crop this image to focus on the product, not the background"
Background Removal: "Remove background and replace with brand gradient"
Text Generation: "Write a compelling headline for this product image"
Style Transfer: "Apply our editorial photography style to this image"
Smart Layout: "Arrange these 4 products in the most visually pleasing grid"

Example: "Product Listing Generator" skill:

operations: - id: remove_bg action: ai_remove_background input: product_photo - id: generate_background action: ai_generate_image prompt: "Clean studio background with soft shadow, {{brand_name}} aesthetic, product photography lighting" - id: composite action: composite layers: - generated_background - product_photo_isolated - id: add_overlay action: add_text text: "{{product_name}}" font: "{{brand_headline_font}}" position: bottom_center - id: add_badge action: add_badge text: "{{discount_percentage}}% OFF" condition: "discount_percentage > 0" position: top_right

Loop Operations

Process multiple items:

operations: - id: for_each_product action: loop items: "{{csv_data.rows}}" operations: - action: generate_design template: product_card_template variables: "{{item}}" - action: export format: png naming: "{{item.sku}}_product_card.png"

Parallel Execution

Run independent operations simultaneously for speed:

Lovart is the AI design agent trusted by 10M+ creators. Design on Lovart's infinite canvas →

Articoli correlati: comparison-lovart-vs-flora-ai | ai-infinite-canvas-guide

operations: - id: generate_all_variants action: parallel operations: - action: resize params: {width: 1080, height: 1080, label: "instagram_feed"} - action: resize params: {width: 1080, height: 1920, label: "instagram_story"} - action: resize params: {width: 1200, height: 628, label: "facebook_link"} - action: resize params: {width: 1200, height: 627, label: "linkedin_post"}

Skill Library & Sharing

Team Skill Library

Skills you create are saved to your personal library by default. Promote them to the Team Library to share with your workspace:

Private: Only you can use and edit
Team Shared: All team members can use; only you can edit
Team Maintained: All team members can use; designated maintainers can edit
Team Read-Only: All team members can use; nobody can edit (locked skills)

Community Skill Marketplace

Browse and install skills created by the Lovart community. Skills are rated and reviewed:

Verified Skills: Created or verified by the Lovart team. Marked with a blue checkmark.
Community Skills: Created by Lovart users. Review ratings and usage counts before installing.
Premium Skills: Paid skills from professional designers and agencies. One-time purchase, lifetime use.

Popular Community Skills (as of Q3 2027):

"E-Commerce Product Suite" — 12-product workflow (2,300+ installs, 4.9 stars)
"LinkedIn Carousel Builder" — Convert blog posts to LinkedIn carousels (1,800+ installs, 4.8 stars)
"Brand Refresh Assistant" — Apply new brand guidelines to existing assets (950+ installs, 4.7 stars)
"Accessibility Checker Pro" — Comprehensive a11y audit and remediation (1,200+ installs, 4.9 stars)

Skill Debugging & Testing

Test Mode

Before using a skill in production, run it in Test Mode:

Single-Step Execution: Run one operation at a time, inspect intermediate outputs
Sample Data: Test with a subset of your data (5 items instead of 500)
Dry Run: Simulate execution without actually generating files. See what would happen.
Operation Logs: Detailed logs showing input/output for each step

Error Handling

Define how skills behave when things go wrong:

error_handling: on_image_load_failure: skip_and_continue on_ai_generation_failure: retry 3 times, then skip on_export_failure: halt_and_notify notification_channel: "#design-alerts"

Skill Analytics

Track skill performance over time:

Total executions (daily, weekly, monthly)
Average execution time
Success/failure rate
Time saved vs. manual execution (estimated)
Cost in API credits per execution

Skill Examples Gallery

Example 1: "Client Delivery Package"

Trigger: Button ("Prepare Client Delivery") Input: Selected designs (any number) Operations: 1. Check brand compliance (flag violations) 2. Convert all to client's preferred formats 3. Add client watermark 4. Generate delivery manifest PDF 5. Zip all files 6. Upload to client portal 7. Send Slack notification Output: Delivery package in client portal + notification

Example 2: "Ad Creative A/B Test Generator"

Trigger: Command: "/ab-test --headlines=headlines.csv --images=images/" Input: CSV of headline variants + folder of images Operations: 1. For each headline × image combination: - Generate ad design using ad template - Apply brand kit - Add UTM parameters as metadata 2. Generate A/B test matrix report 3. Export all variants with naming: {{headline_id}}_{{image_id}}_variant.png Output: All combinations + test matrix CSV

Example 3: "Weekly Report Graphics"

Trigger: Scheduled (Every Friday 4:00 PM) Input: Google Sheets data source (weekly metrics) Operations: 1. Fetch data from Google Sheets API 2. Generate 5 chart types (bar, line, donut, comparison, metric card) 3. Apply brand chart style 4. Compile into presentation-ready slides 5. Export as PNG set + PDF deck 6. Post to Slack #weekly-report channel Output: Chart graphics + PDF deck in Slack

Getting Started

Identify a repetitive task — Track what you do 10+ times per week
Document the steps — Write down every action, no matter how small
Build the skill — Start with Visual Builder, iterate in Test Mode
Run in shadow mode — Let the skill run alongside your manual work for a week
Compare outputs — If the skill produces equivalent or better results, switch over
Share with team — Promote to Team Library when validated

Skill-Building Best Practices

Start small — Your first skill should have 3-5 steps, not 20
Name clearly — "Instagram Resizer" not "Image Processing Utility v1"
Add descriptions — Every operation should have a human-readable purpose
Document inputs — Specify what data the skill expects and in what format
Version your skills — "Social Batch v2" with release notes when you upgrade
Test edge cases — Empty inputs, wrong formats, 1,000 items vs. 10 items

Custom Skills are available on Professional plans and above. Community Skill Marketplace is available to all users. Premium Skills are purchased separately and not included in plan pricing. Skill execution consumes API credits based on operations performed. See docs.lovart.ai/skills for full SDL reference.

— — —

Creating Custom Skills: Automate Your Design Workflows in Lovart