seo-audit

A Claude Code skill that runs a repo-wide, read-only sweep for missing or incomplete technical-SEO practices — a fixed checklist of best practices (search-engine verification, crawlable pagination, responsive images, canonical tags, structured data, meta title/description, heading hierarchy, alt text, robots.txt/sitemap health, hreflang, redirect chains, HTTPS/mixed content, Core Web Vitals signals, URL structure, JS-render/prerender coverage, Open Graph/Twitter tags, per-page robots meta), rather than hunting for code that's actively wrong.

Framework-agnostic by design — the checklist categories describe what to check and why, with discovery hints for finding the relevant implementation across common stacks (React/CRA, Next.js, Vite, plain server-rendered apps, etc.), not hardcoded paths.

From your project's repo root:

curl -sL https://raw.githubusercontent.com/tushar33/seo-audit-skill/main/install.sh | bash

This copies skills/seo-audit/{seo-audit.md,categories.md} into your project and symlinks the skill into .claude/commands/seo-audit.md so Claude Code picks it up as /seo-audit.

1. Copy skills/seo-audit/categories.md and skills/seo-audit/seo-audit.md into that project's skills/seo-audit/ directory, unchanged.
2. Symlink seo-audit.md into the project's .claude/commands/ directory so Claude Code resolves it as /seo-audit:

   ln -s ../../skills/seo-audit/seo-audit.md .claude/commands/seo-audit.md

3. Run /seo-audit in that project. Since it has no project-checklist.md yet, the first run treats every category as unaudited, uses categories.md's discovery hints to find the equivalent implementation in that codebase, and creates a fresh project-checklist.md there as it goes.

• categories.md — generic, portable, framework-agnostic. Defines each technical-SEO category, why it matters, and how to discover the relevant implementation in an unfamiliar codebase. No hardcoded paths. Safe to copy verbatim into any project.
• seo-audit.md — the skill logic itself (also generic/portable): how to run the audit, verify findings before reporting them, and keep a per-project checklist up to date across runs.
• project-checklist.md — generated per-project, not included here. Each project that installs this skill gets its own instance with real file paths, current state, and a fast re-verification recipe per category. Never copy one project's checklist into another.

/seo-audit                      # every category, using project-checklist.md's fast path + discovery for gaps
/seo-audit SEO-01               # one category only (e.g. search-engine verification)
/seo-audit add: <describe a new universal technical-SEO practice to track>

Every finding is re-verified against current code (a checklist entry records a last-known state, but code moves) and reported as STILL_OPEN, PARTIALLY_RESOLVED, RESOLVED, or NOT_APPLICABLE. The audit changes no code — fixes go through explicit approval or your own follow-up process.

Each run also saves its full report — every category's status plus the summary table — to skills/seo-audit/reports/<YYYY-MM-DD>.md in the project being audited, so a given run's complete output is preserved as its own shareable file, separate from project-checklist.md's cumulative last-known-state tracking.

• New universal practice (applies to any project, no project-specific paths) → add it to categories.md.
• A finding for a specific project → belongs in that project's own project-checklist.md, not here.
• Wrong code already causing bugs (not a missing feature) is a different kind of check — this skill only tracks gaps against the checklist, not active bugs.

See CONTRIBUTING.md for ground rules before opening a PR.

MIT