Article SEO Audit

Article SEO Audit is an Agent Skill that turns any public article or blog post URL into a thorough, scored on-page SEO report with prioritized, copy-paste-ready fixes. You give it a URL (and optionally a target keyword), and the skill fetches the page, extracts every on-page SEO signal, scores each against a documented rubric, and writes an actionable report led by the highest-impact changes.

What is Article SEO Audit?

It is a free, open-source skill that packages a repeatable on-page SEO workflow for AI agents. A standard-library Python extractor pulls the raw signals from a page; the agent then evaluates them against a 15-dimension rubric and produces a 0–100 score, a letter grade, and a fix list. It covers on-page and content SEO — the signals visible in the page HTML — and is explicit about what it does not measure rather than guessing.

What it does

• Extracts the full on-page signal set: title and length, meta description, robots and X-Robots-Tag, canonical, lang, viewport, hreflang, the full heading outline with H1 count and level nesting, content word count, link profile (internal/external/nofollow/empty-anchor), image alt-text coverage, JSON-LD structured data types, Open Graph and Twitter Card completeness, published/modified dates, readability (Flesch reading ease and grade level), and keyword placement and density.
• Scores 15 dimensions — title, meta, URL slug, headings, content depth, keyword optimization, readability, images, links, structured data, social cards, crawlability, freshness, E-E-A-T, and SERP-feature readiness — each marked Pass, Needs work, or Fail.
• Produces a prioritized report: a verdict, the 3–6 highest-impact fixes (each with what is wrong, why it matters, and a concrete fix), a full dimension table, quick wins vs larger efforts, and a raw-data appendix.

How it works

Run the extractor from the skill directory — python3 scripts/seo_extract.py "<URL>" --keyword "<target keyword>" — which fetches the page with a browser user-agent, follows redirects, and prints one JSON object with every signal. The agent reads that JSON, applies the thresholds in references/audit-criteria.md, computes the weighted score, and writes the report using references/report-template.md. If a site blocks the fetch (403, Cloudflare) or renders content with JavaScript, the skill falls back to WebFetch or user-supplied HTML and states which method produced the data.

Who it's for

Content marketers, SEO specialists, bloggers, and editorial teams who want a fast, consistent on-page review before publishing or as a refresh pass — and anyone using an AI agent like Claude Code who wants SEO feedback grounded in extracted page data instead of generic advice.

What you can build

• A pre-publish SEO checklist that runs against a draft URL
• Head-to-head audits comparing your page to a competitor ranking for the same keyword
• Rewritten title tags, meta descriptions, and heading structures handed back as ready-to-paste fixes
• A repeatable scoring system to track a page's on-page SEO over time

Why it matters

Most "SEO checks" are either shallow checklists or paid tools that hide the reasoning. This skill is transparent and grounded: every score traces to an extracted signal and a documented threshold, every fix is specific, and it refuses to fabricate metrics it cannot measure — no invented PageSpeed scores, backlink counts, or keyword volumes. It is honest about its scope (on-page only, not backlinks or Core Web Vitals), which makes the output something you can actually act on.

Install with the skills CLI

npx skills add lokeshkapoor/skills --skill article-seo-audit

Or clone the skill folder manually

git clone https://github.com/lokeshkapoor/skills
# the skill lives in skills/article-seo-audit

Run an audit

Once installed, ask your agent to audit a URL, or run the extractor directly:

python3 scripts/seo_extract.py "https://example.com/your-article" --keyword "your target keyword"

The agent reads the extracted signals, scores them against references/audit-criteria.md, and writes the report using references/report-template.md.