Full-Site SEO Audit
Crawl a website and turn SEO signals into a scored, prioritized action plan.
The skill specifies respecting robots.txt, crawl limits, optional credentials, and conditional API integrations; repository security material further describes SSRF controls, credential permissions, and local isolation. However, SKILL.md does not require user confirmation before crawling, fully disclose all data flows or external effects, or define rollback for persisted artifacts, so points are deducted.
The process defines timeouts, concurrency, redirect limits, abnormal URL/rate-limit/timeout handling, and output contracts; the repository also contains CI and related tests. Static evidence does not reproduce the 500-page crawl, parallel delegation, renderer, or optional integrations end to end, and recovery diagnostics remain limited, so the score is capped conservatively.
Audience scenarios, trigger phrases, URL input, specialist coverage, conditional delegation, and outputs are relatively clear. Non-fit boundaries, Chinese-language behavior, and mainland-China reachability are not specified; core operation may depend on external sites, Playwright, Google/API credentials, or optional MCP services, so points are deducted.
Front matter, version, license, staged process, crawl configuration, output artifacts, report structure, and error table are documented; README, LICENSE, CI, and security policy add installation and maintenance signals. Dependency setup, parameter stability, changelog/update path, and explicit maintenance ownership are mostly outside the selected SKILL.md, limiting maintainability.
The skill defines structured JSON, Markdown, action-plan, findings, and optional PDF deliverables, with business detection, conditional specialists, prioritization, and broad SEO coverage. Static review cannot verify report correctness, aggregation quality, specialist completeness, or PDF usability; outputs still require human review, so the score remains at the static ceiling.
CI, committed tests, static instruction guards, security regression tests, and a sample report provide an auditable basis. There is no independently reproducible, revision-pinned end-to-end execution record for this full audit path, and README performance/result claims cannot be independently confirmed from the supplied static evidence.
- Auditing arbitrary user-supplied sites causes external HTTP requests; confirm target authorization and data-handling boundaries, and review redirect, renderer, and third-party resource risks.
- Google, DataForSEO, Moz, Bing, and GA4 integrations require credentials or external services; unavailable or restricted integrations must be reported as partial coverage rather than inferred.
- The skill writes to {domain}-audit/ and may create PDFs, screenshots, and structured data; confirm output location, redaction requirements, and handling of leftover artifacts after failure.
- Chinese output behavior and mainland-China network reachability are unspecified; validate external APIs, target-site access, and Playwright dependencies before deployment to Chinese users.
What does this skill do, and when should you use it?
This is the full-site audit skill inside the claude-seo repository. It crawls up to 500 pages, detects the business type, and can delegate analysis to multiple SEO specialist subagents. Its coverage includes technical SEO, content quality, structured data, performance, images, and AI search readiness, combined into a 0–100 SEO Health Score. Results are persisted as Markdown, JSON, screenshots, and optionally PDF or HTML reports.
It renders the homepage with scripts/render_page.py to collect raw HTML, rendered HTML, extracted text, SPA status, and accessibility data when needed. It analyzes homepage signals for business type, respects robots.txt, follows internal links for up to 500 pages, allows up to three redirect hops, uses a 30-second page timeout, five concurrent requests, and a one-second delay between requests. It checks robots.txt, sitemaps, canonicals, Core Web Vitals, security headers, E-E-A-T, readability, thin content, Schema, internal linking, images, and AI search readiness. When available, it delegates work to technical, content, schema, sitemap, performance, visual, GEO, local SEO, SXO, and other specialist subagents. It then produces an SEO Health Score, prioritized action plan, category findings, and structured audit data.
- An SEO agency needs a repeatable full-site audit for a client and a Critical-to-Low prioritized action plan.
- An in-house SaaS, publisher, or e-commerce SEO lead wants a pre-review check of technical issues, content quality, structured data, and Core Web Vitals.
- A freelance consultant wants a concrete health score and evidence-backed findings during initial client discovery.
- A site operator needs explicit reporting when a URL is unreachable, robots.txt restricts crawling, requests are rate-limited, or a large crawl times out.
What are this skill's strengths and limitations?
- The crawl limits, concurrency, scoring weights, priority definitions, and output envelope are explicitly specified.
- It supports SPA rendering, crawling up to 500 pages, robots.txt compliance, and multi-category specialist analysis.
- The core audit can run without Google API credentials or MCP extensions.
- Its error handling covers unreachable URLs, robots.txt restrictions, 429 responses, and large-site timeouts.
- The workflow relies on Claude Code slash-command and subagent-delegation behavior, so other clients require adaptation.
- Without Google APIs, Core Web Vitals are mainly lab estimates and indexation is inferred from page-level signals.
- Heavy hydration, scroll-triggered content, or interaction-dependent loading can produce noisy findings.
- The source provides no standalone installation steps or independent test results for this individual skill.
How do you install this skill?
The README documents installation for the complete claude-seo collection, not for this skill alone. On Unix/macOS/Linux: git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git, then run bash claude-seo/install.sh. On Windows PowerShell: git clone --depth 1 https://github.com/AgriciDaniel/claude-seo.git, then run powershell -ExecutionPolicy Bypass -File claude-seo\install.ps1. The source does not document a standalone installation procedure for skills/seo-audit/SKILL.md.
How do you use this skill?
In Claude Code, run /seo audit https://example.com, or provide a URL in a request containing “audit,” “full SEO check,” “analyze my site,” or “website health check.” Results are written under {domain}-audit/, including FULL-AUDIT-REPORT.md, ACTION-PLAN.md, audit-data.json, category findings, and screenshots when available. The source also documents optional PDF generation with scripts/google_report.py.
How does this skill compare with similar options?
Compared with Screaming Frog, this skill emphasizes conversational synthesis, specialist delegation, and action planning inside Claude Code; the README states that Screaming Frog crawls deeper and faster at the link-graph level. Compared with Ahrefs Site Audit, it does not include Ahrefs' proprietary backlink index, although an optional MCP extension can connect Ahrefs. Its core focus is local-file reporting, AI search readiness, and verifiable recommendations.