web-research

Web Research Protocol

When to Use

Use this skill when the user wants to:

• Audit a website (performance, SEO, security, accessibility, tech stack)
• Compare competitors or analyze multiple sites
• Conduct structured web research with evidence
• Review site quality, trust signals, or conversion flows
• Extract structured data across multiple pages

Protocol: Acquire → Normalize → Analyze

Every research task follows three phases:

Phase 1: Acquire

Connect to each target page and extract comprehensive data using smart.extractPage():

// Connect first
await connect_tab({ url: "https://example.com", background: true })

// Extract everything in one call
const page = await smart.extractPage()
// Returns: { capture, performance, security, fonts, meta, accessibility, mobileReadiness, gaps }

For visual evidence, add smart.scrollCapture():

const visuals = await smart.scrollCapture({ maxSections: 5 })
// Returns: { sectionCount, sections: [{ index, scrollY, screenshot }] }

Wait for dynamic content with smart.waitForIdle() instead of sleep():

const idle = await smart.waitForIdle(5000)
// Returns: { status: 'idle' | 'timeout' }

Phase 2: Normalize

Structure evidence into canonical per-page records. Each page record should include:

const record = {
  url: page.capture.url,
  title: page.capture.title,
  framework: page.capture.framework,
  network: page.capture.network,        // { total, failed, byType, errors }
  console: page.capture.console,        // { total, errors, warnings }
  cookies: page.capture.cookies,        // { total, names }
  dom: page.capture.dom,                // { nodeCount, forms, links, images, inputs }
  performance: page.performance,        // LCP, CLS, FCP, Web Vitals
  security: page.security,             // TLS, cert, mixed content
  fonts: page.fonts,                   // declared + computed
  meta: page.meta,                     // OG tags, structured data, headings, nav links
  accessibility: page.accessibility,   // node count, landmarks, images without alt, heading structure
  mobileReadiness: page.mobileReadiness, // viewport meta, media queries, overflow
  gaps: page.gaps,                     // what data failed — check before trusting null fields
}