🍔 FAT Agent — Fix, Audit, Test

A Claude plugin that acts as your post-launch QA engineer.

FAT Agent systematically audits deployed websites for SEO, performance, security, accessibility, and content issues — then walks you through fixing every one.

---

What It Does

After you deploy a site, say "run FAT agent" and it will:

1. Gather context — Asks smart questions about your site, stack, and critical user flows
2. Audit — Runs 8 automated check categories against your live URL
3. Report — Generates a prioritised punch list (P0 Critical → P3 Nice-to-have)
4. Fix — Offers to generate code fixes for every issue found
5. Re-test — After you redeploy, verifies the fixes are live

Audit Categories

Category	Automated?	What It Checks
🌐 Availability & Response	✅	HTTP status, redirects, response headers, caching
🔍 SEO Essentials	✅	Title, meta, headings, OG tags, structured data, sitemap, robots.txt
⚡ Performance	✅	HTML size, render-blocking scripts, lazy loading, resource hints
🔒 Security Headers	✅	HSTS, CSP, X-Frame-Options, Referrer-Policy, Permissions-Policy
♿ Accessibility	Partial	Alt text, labels, landmarks, skip links + targeted user questions
🧪 Functional Checks	👤 User	Forms, navigation, mobile, 404 page, integrations
📝 Content & Legal	Partial	Placeholder text, privacy policy, copyright year
📊 Analytics & Tracking	✅	GA4, GTM, Facebook Pixel, Plausible, Hotjar, Fathom, Umami, Mixpanel, PostHog + 10 more

---

Installation

Claude Code Plugin (Recommended)

claude plugins add https://github.com/spruikco/fat-agent-skill

This installs the FAT Agent plugin with the /fat-audit slash command.

Claude Code (Manual)

git clone https://github.com/spruikco/fat-agent-skill ~/.claude/skills/fat-agent

Claude Code reads SKILL.md automatically and activates the skill when it detects trigger phrases.

Then in any conversation:

You: Run FAT agent on https://mysite.com
You: Audit my site
You: I just deployed — is everything working?
You: Post-launch check on https://example.com
You: /fat-audit https://example.com

Claude.ai (Projects)

1. Create a new Project in Claude.ai
2. Upload plugins/fat-agent/skills/fat-agent/SKILL.md as a project file — this is the core instruction set Claude follows
3. Upload the reference files you want available:
   • plugins/fat-agent/references/security-headers.md
   • plugins/fat-agent/references/seo-checklist.md
   • plugins/fat-agent/references/accessibility-guide.md
   • Any relevant plugins/fat-agent/references/platform-fixes/*.md for your hosting platform
   • Any relevant plugins/fat-agent/references/framework-fixes/*.md for your tech stack
4. Start a conversation and say "audit my site" or "run FAT agent"

Note: The Python scripts (analyse-html.py, calculate-score.py) are designed for Claude Code, which can execute them directly. Claude.ai performs the same checks conversationally using web_fetch.

What Happens When You Trigger It

1. Phase 0 — Context — FAT Agent asks for your live URL, site type, tech stack, and hosting platform
2. Phase 1 — Audit — Fetches your URL, runs 9 check categories (SEO, security, accessibility, performance, analytics, content, functional, platform-specific), asks targeted yes/no questions for things that can't be automated
3. Phase 2 — Fix — Generates a prioritised punch list (P0 Critical → P3 Low) with specific code/config fixes tailored to your stack and hosting platform
4. Phase 3 — Test — After you redeploy with fixes, re-fetches the URL and verifies each issue is resolved

Fix suggestions are loaded on-demand from the references/platform-fixes/ and references/framework-fixes/ directories based on what you told it in Phase 0. A Next.js site on Vercel gets different fix code than a WordPress site on Apache.

Works With Any Hosting Platform

FAT Agent is platform-agnostic. It audits the live URL regardless of where it's hosted:

• Netlify, Vercel, Cloudflare Pages
• AWS (S3, CloudFront, Amplify, EC2)
• DigitalOcean, shared hosting (cPanel, Plesk)
• Self-hosted (Nginx, Apache)
• WordPress hosting (WP Engine, Kinsta, etc.)
• Any platform that serves a URL

---

Project Structure

fat-agent-skill/                          # Marketplace root
├── .claude-plugin/
│   └── marketplace.json                  # Marketplace manifest
├── plugins/
│   └── fat-agent/                        # The plugin
│       ├── .claude-plugin/
│       │   └── plugin.json               # Plugin manifest
│       ├── skills/
│       │   └── fat-agent/
│       │       └── SKILL.md              # Core skill instructions
│       ├── commands/
│       │   └── fat-audit.md              # /fat-audit slash command
│       ├── scripts/
│       │   ├── analyse-html.py           # HTML analysis helper
│       │   ├── calculate-score.py        # Scoring calculator (SEO, Security, A11y, Perf, FAT)
│       │   ├── generate-badge.py         # SVG badge generator for READMEs
│       │   ├── track-history.py          # Historical audit tracking
│       │   └── test_fat_agent.py         # Full test suite (285 tests)
│       ├── references/
│       │   ├── security-headers.md       # Security header reference
│       │   ├── seo-checklist.md          # Extended SEO criteria
│       │   ├── accessibility-guide.md    # WCAG 2.1 quick reference
│       │   ├── performance-budgets.md   # Performance budget configuration
│       │   ├── ci-cd-integration.md     # CI/CD integration examples
│       │   ├── platform-fixes/           # Hosting platform config guides
│       │   │   ├── netlify.md
│       │   │   ├── vercel.md
│       │   │   ├── cloudflare-pages.md
│       │   │   ├── apache.md
│       │   │   ├── nginx.md
│       │   │   ├── wordpress.md
│       │   │   └── aws.md
│       │   └── framework-fixes/          # Framework-specific fix patterns
│       │       ├── nextjs.md
│       │       ├── astro.md
│       │       ├── sveltekit.md
│       │       ├── nuxt.md
│       │       ├── gatsby.md
│       │       ├── wordpress.md
│       │       └── static-html.md
│       ├── evals/
│       │   └── evals.json                # Test cases for skill validation
│       ├── assets/
│       │   ├── fat-agent-badge-icon.png
│       │   └── social-preview.png
│       └── README.md                     # Plugin documentation
├── fat-badge.svg                         # Generated FAT score badge
├── README.md                             # This file
├── CLAUDE.md                             # Project conventions for Claude Code
├── LICENSE                               # MIT License
├── .gitignore
└── .github/
    └── LLM-BRIEF.md                     # Project brief for LLM continuation

---

Usage Examples

Basic Audit

User: Run FAT agent on https://mysite.com
Claude: Ready to run a FAT audit! I just need a few details...

Post-Deploy Check

User: I just deployed. Is everything working?
Claude: Let me run a FAT audit on your site to check...

Targeted Audit

User: Can you check the SEO on my new landing page?
Claude: I'll focus the FAT audit on SEO — fetching your page now...

Slash Command

User: /fat-audit https://mysite.com
Claude: Running FAT Agent audit on https://mysite.com...

---

The FAT Report

Issues are prioritised with clear labels:

Priority	Label	Meaning
🔴 P0	Critical	Site is broken, inaccessible, or insecure
🟠 P1	High	Significant SEO, performance, or UX impact
🟡 P2	Medium	Best practice violations, minor issues
🟢 P3	Low	Nice-to-haves, polish items

Each finding includes:

• What's wrong — One-line description
• Why it matters — Impact explanation
• How to fix — Specific code/config changes
• Effort — ⚡ 5 min, 🔧 30 min, or 🏗️ 1+ hour

---

Scoring

FAT Agent generates scores across four dimensions:

• SEO Score (0-100) — Based on meta tags, headings, structured data, sitemap, etc.
• Security Score (0-100) — Based on presence and correctness of security headers
• Accessibility Score (0-100) — Based on automated checks + user-reported items
• Performance Score (0-100) — Based on HTML size, render-blocking scripts, images, fonts, lazy loading
• Overall FAT Score — Weighted composite (SEO 30%, Security 25%, A11y 30%, Perf 15%)

FAT Badge

Generate shields.io-style SVG badges from your audit scores:

# Overall FAT badge (grade + score)
python plugins/fat-agent/scripts/analyse-html.py page.html | python plugins/fat-agent/scripts/calculate-score.py | python plugins/fat-agent/scripts/generate-badge.py --output badge.svg

# Category badges
python plugins/fat-agent/scripts/generate-badge.py scores.json --category seo --output seo-badge.svg
python plugins/fat-agent/scripts/generate-badge.py scores.json --category security --output security-badge.svg
python plugins/fat-agent/scripts/generate-badge.py scores.json --category accessibility --output a11y-badge.svg
python plugins/fat-agent/scripts/generate-badge.py scores.json --category performance --output perf-badge.svg

# Flat-square style
python plugins/fat-agent/scripts/generate-badge.py scores.json --style flat-square --output badge.svg

Then embed in your README:

![FAT Score](./badge.svg)

---

Customisation

Adding Check Categories

Edit plugins/fat-agent/skills/fat-agent/SKILL.md to add new audit sections. Follow the existing pattern:

1. Add the check to the appropriate phase
2. Specify whether it's automated or user-prompted
3. Define the priority level for findings
4. Add fix templates

Extending References

Drop additional .md files in plugins/fat-agent/references/ and reference them from SKILL.md.

---

Contributing

PRs welcome! All roadmap items are now implemented:

• More comprehensive accessibility checks (ARIA validation, tabindex, autoplay, zoom, tables, SVGs, iframes)
• Performance budget configuration (.fat-budget.json + --budget flag)
• CI/CD integration examples (GitHub Actions, Netlify, Vercel, GitLab CI, generic shell)
• Additional analytics provider detection (17+ new providers + 3 consent platforms)
• FAT Badge generator (SVG score badge for READMEs)
• Historical audit tracking (track-history.py — save, show, diff, trend)
• Competitive analysis mode (side-by-side comparison in SKILL.md)
• Hardened SEO checks (thin content, anchor text, link audit, URL structure, image filenames, duplicate OG, canonical validation, CWV integration)

---

Credits

Built by Spruik Co — Digital Marketing & SEO Consultancy.

Designed as a Claude Agent Skill for post-launch quality assurance.

---

License

MIT — see LICENSE for details.