How Skills Activate
Skills activate automatically when a user’s request matches the skill’s description. The description in the YAML frontmatter contains trigger phrases that AI agents use to determine relevance.You don’t explicitly invoke skills - they activate automatically when your request matches their trigger phrases. This creates a natural, conversational experience.
Trigger Phrases by Skill
Each Web Quality Skill includes specific trigger phrases in its description:1. Web Quality Audit
- Trigger Phrases
- Description
- What It Does
- “audit my site”
- “review web quality”
- “run lighthouse audit”
- “check page quality”
- “optimize my website”
2. Performance
- Trigger Phrases
- Description
- What It Does
- “speed up my site”
- “optimize performance”
- “reduce load time”
- “fix slow loading”
- “improve page speed”
- “performance audit”
3. Core Web Vitals
- Trigger Phrases
- Description
- What It Does
- “improve Core Web Vitals”
- “fix LCP”
- “reduce CLS”
- “optimize INP”
- “page experience optimization”
- “fix layout shifts”
4. Accessibility
- Trigger Phrases
- Description
- What It Does
- “improve accessibility”
- “a11y audit”
- “WCAG compliance”
- “screen reader support”
- “keyboard navigation”
- “make accessible”
5. SEO
- Trigger Phrases
- Description
- What It Does
- “improve SEO”
- “optimize for search”
- “fix meta tags”
- “add structured data”
- “sitemap optimization”
- “search engine optimization”
6. Best Practices
- Trigger Phrases
- Description
- What It Does
- “apply best practices”
- “security audit”
- “modernize code”
- “code quality review”
- “check for vulnerabilities”
Example Activations
Here’s how different user requests activate skills:Example 1: Single Skill
Example 1: Single Skill
User: “Can you help speed up my site?”Activated Skill: PerformanceWhy: The phrase “speed up my site” directly matches the performance skill’s trigger phrases.Agent Response:
- Analyzes performance bottlenecks
- Checks performance budget adherence
- Reviews critical rendering path
- Provides specific optimization recommendations
Example 2: Multiple Skills
Example 2: Multiple Skills
User: “Audit this page for accessibility and SEO issues”Activated Skills: Accessibility + SEOWhy: Request mentions both “accessibility” and “SEO” explicitly.Agent Response:
- Runs accessibility audit using WCAG guidelines
- Performs SEO analysis for technical and on-page factors
- Provides separate reports for each area
- Prioritizes issues by severity
Example 3: Orchestrated Skills
Example 3: Orchestrated Skills
User: “Run a complete quality audit on my website”Activated Skill: Web Quality Audit (which orchestrates others)Why: “quality audit” matches the web-quality-audit skill.Agent Response:
- Loads web-quality-audit skill as orchestrator
- References performance skill for performance checks
- References accessibility skill for a11y checks
- References SEO skill for search optimization
- References best-practices skill for security/code quality
- Provides comprehensive report across all categories
Example 4: Specific Metric
Example 4: Specific Metric
User: “My LCP is 4.5 seconds, help me fix it”Activated Skill: Core Web VitalsWhy: “LCP” is a specific trigger phrase for core-web-vitals skill.Agent Response:
- Focuses specifically on LCP optimization
- Identifies common LCP issues (server response, render blocking, etc.)
- Provides LCP-specific code examples
- May load LCP.md reference for deeper guidance
Skill Orchestration
The Web Quality Audit skill demonstrates orchestration - using multiple skills together:How Orchestration Works
High-Level Analysis
Web quality audit skill provides structure:
- Audit categories (Performance, Accessibility, SEO, Best Practices)
- Severity levels (Critical, High, Medium, Low)
- Output format template
Deep Dives
Agent loads specific skills as needed:
- Performance skill for detailed performance analysis
- Core Web Vitals skill when LCP/INP/CLS issues found
- Accessibility skill for WCAG compliance
- SEO skill for search optimization
- Best Practices skill for security/code quality
Best Practices for Invoking Skills
1. Be Specific When Possible
- Less Specific
- More Specific
“Make my site better”Result: May activate web-quality-audit for general analysis, but vague.
2. Use Natural Language
You don’t need to memorize exact trigger phrases. Natural language works:
- “How can I make my images load faster?” → Performance skill
- “Is my site accessible to screen readers?” → Accessibility skill
- “Why isn’t Google finding my pages?” → SEO skill
3. Combine Multiple Concerns
4. Provide Context
- Without Context
- With Context
“Fix performance”Result: Generic performance analysis.
5. Reference Specific Metrics
Using metric names activates precise skills:| Mention | Activates | Focus |
|---|---|---|
| ”LCP is slow” | Core Web Vitals | Largest Contentful Paint optimization |
| ”Fix layout shifts” | Core Web Vitals | CLS-specific solutions |
| ”Buttons feel laggy” | Core Web Vitals | INP optimization |
| ”Contrast issues” | Accessibility | WCAG contrast compliance |
| ”Missing meta descriptions” | SEO | On-page SEO |
Multi-Skill Workflows
Some workflows benefit from using multiple skills in sequence:Workflow 1: Pre-Launch Checklist
Workflow 2: Specific Problem Solving
Activation Priority
When multiple skills could apply, agents prioritize based on:- Explicit mentions - Direct skill names or specific metrics
- Specificity - More specific skills over general ones
- Context - Recent conversation history
- Orchestration - Comprehensive skills that reference others
Example: Ambiguous Request
Example: Ambiguous Request
User: “Optimize my homepage”Analysis:
- Could match: Performance, SEO, Accessibility, Web Quality Audit
- Most likely activation: Web Quality Audit (most comprehensive)
- Why: “Optimize” is vague, so comprehensive audit provides best coverage
- “Optimize my homepage speed” → Performance skill
- “Optimize my homepage for Google” → SEO skill