Nebula Landing Page Audit

Nebula Landing Page Audit Skill

Use this skill to diagnose why a landing page isn't converting, especially when paid traffic is already running. It runs an evidence-grade 9-signal audit and returns a prioritized fix list — not opinions.

When to use this skill

• "Why aren't my ads converting?"
• "Audit this landing page"
• "Find conversion leaks on [URL]"
• "Compare my page to a competitor's"
• "What's wrong with this page's CTA / headline / social proof?"

MCP Server

This skill is backed by a live remote MCP server:

https://mcp.nebulacomponents.com/mcp

The server requires no API key. All audits are free.

Available Tools

run_audit(url: str)

Run a full evidence-grade audit against any public URL.

Returns: Score (0–10), letter grade (A–F), and structured findings. Each finding includes:

• label — the signal being tested
• quadrant — Quick Win / Major Project / Fill In / Hard Slog (effort × impact)
• issue — what was observed
• fix — the specific remediation
• evidence — measured value, required standard, delta
• impact / effort — numeric scores (0–10)
• confidence — definitive / high / contextual

Example:

run_audit("https://example.com")

Example output:

## Audit: https://example.com
Score: 6.2/10  Grade: B

Findings (4):

**Message Match** [Quick Win]
  Issue:    Ad headline "Stop wasting ad spend" doesn't match page headline "We help businesses grow"
  Fix:      Mirror the ad's outcome language in the H1
  Measured: "We help businesses grow"
  Required: Outcome-specific headline matching ad copy
  Delta:    Mismatch detected

**CTA Clarity** [Quick Win]
  Issue:    Two competing CTAs above the fold with equal visual weight
  Fix:      Make one CTA dominant; reduce contrast on secondary action
  Confidence: definitive

compare_audits(url_a: str, url_b: str)

Audit two pages and compare them side by side.

Use this to benchmark your page against a competitor's, or compare a variant before committing to a redesign.

Example:

compare_audits("https://yourpage.com", "https://competitor.com")

get_audit(audit_id: str)

Retrieve a completed audit by its UUID. Use this to re-examine a previous audit without re-running it.

Example:

get_audit("a1b2c3d4-e5f6-7890-abcd-ef1234567890")

recent_audits(limit: int = 10)

List the most recent audits in the Nebula database. Returns URL, score, grade, and timestamp. Useful for surveying what's been audited.

Scoring Reference

Signal Definitions

Usage patterns

Basic audit:

Please audit this page: https://example.com/pricing

Competitive analysis:

Compare our landing page to our top competitor:
- Ours: https://oursite.com/lander
- Theirs: https://competitor.com/landing

Pre-launch check:

We're about to run Google Ads to https://example.com/new-campaign.
Run an audit and tell me what to fix before we spend money.

Quick Win filter:

Audit https://example.com and list only the Quick Win findings
(high impact, low effort) so we can ship fixes today.

MCP Configuration

Claude Desktop

Add to claude_desktop_config.json:

{
  "mcpServers": {
    "nebula": {
      "command": "npx",
      "args": [
        "mcp-remote",
        "https://mcp.nebulacomponents.com/mcp"
      ]
    }
  }
}

Hermes Agent

hermes mcp add nebula --url https://mcp.nebulacomponents.com/mcp

Any MCP-compatible client

Transport: streamable-http
Endpoint:  https://mcp.nebulacomponents.com/mcp
Auth:      None required

Evaluation

Test case 1 — Basic audit

Prompt: Run an audit on https://nebulacomponents.com

Expected:

• Returns score between 0–10
• Returns grade letter
• Returns at least 1 finding with label, issue, fix fields
• Completes in < 30 seconds

Test case 2 — Findings structure

Prompt: Audit https://example.com and list all Quick Win findings

Expected:

• Agent calls run_audit
• Agent filters findings where quadrant == "Quick Win"
• Response includes fix for each finding

Test case 3 — Comparison

Prompt: Compare https://nebulacomponents.com and https://stripe.com/payments

Expected:

• Agent calls compare_audits with both URLs
• Response includes scores for both pages
• Response identifies which page scores higher and why

Test case 4 — Error handling

Prompt: Audit https://localhost/private

Expected:

• Tool returns an error or low-confidence result
• Agent does not hallucinate a score

Notes

• Audits typically complete in 5–15 seconds
• The tool fetches and analyzes the live page at time of call
• Password-protected or Cloudflare-blocked pages may return limited findings
• All audits are free — no rate limits for reasonable usage