v1.3.03 loop iterations

On-Page SEO Auditor

Most pages leave 20-40% of their on-page potential on the table because nobody checked. The On-Page SEO Auditor scores any page against 80+ ranking signals — and is honest about which signals it could and couldn't see in the input you gave it.

What this skill does

On-page SEO is the part you control directly. Unlike backlinks or domain authority, the on-page signals can be fixed in an afternoon and start affecting rankings within days. The audit covers 80+ checks across title and meta, heading structure, content quality, internal/external links, images, URL structure, schema, technical on-page, UX signals, FAQ, and accessibility — every check scored PASS, FAIL, WARN, or N/A with the exact fix.

The N/A column is the part most auditors get wrong, and it matters. A markdown draft with no <head> shouldn't be marked FAIL on the title tag — it hasn't been written yet. A raw view-source of a JS app shouldn't be marked FAIL on H1 and word count — that's an unrendered shell, not a thin page. The skill scopes the input first, runs only the checks the input can legitimately support, and reports the score over assessable checks. A great draft scores like a great draft, not 55% because it has no canonical tag yet. N/A is never a shortcut to a shorter report either — every body-level check on a markdown post is fully scored, because pretending you can't see them is the same trust failure as over-FAILing.

Two checks override everything else: search-intent match and content quality. A page that's mechanically 95% but answers the wrong intent is not a 95% page — a how-to that ranks against listicles will lose. The skill classifies the keyword's intent, states what the page actually delivers in one line, compares them, and makes the intent fix the headline finding when they disagree. Keyword stuffing scores FAIL, not "well-targeted" — more keyword is never better.

A few honesty rules baked in: it won't manufacture a critical issue to fill the section. If the page is strong, "no critical issues" is the correct output and the report is short. It won't soften when a user says "it already ranks, just confirm it's perfect" — current ranking is not proof a page is on-page-maxed. And every finding lands with the specific tag, the specific line, and the specific replacement — never "improve your content quality".

When this triggers

·You wrote a post and want to know what's missing before you publish
·A page isn't ranking and you don't know if the on-page work is the cause
·You're handed a markdown draft and need to audit before the `<head>` exists
·You're auditing a JS-app page and need someone who won't false-FAIL the unrendered shell
·You want a scored report you can hand to a client with specific fixes, not vague advice

Example

Trigger

User pastes a 1,400-word blog post in markdown. Says: 'Target keyword is claude code skills. Will this rank?'

Output

# SEO Audit: Building Your First Claude Code Skill Target keyword: claude code skills INTENT VERDICT: informational/how-to query, post is a how-to. Match — proceed. Score: 47/61 checks passed (77%). 22 checks N/A (markdown draft, no <head> or URL yet). CRITICAL: · Primary keyword absent from first 100 words. Add to intro. · H1 missing entirely (markdown opens with H2). · Internal links: 0. Target 4-6 for 1,400 words. WARNINGS: · Keyword density 0.3% — under-targeted, not over. · No FAQ section. Cluster suggests 4 natural questions. · 3 instances of 'leverage' as a verb. AI slop tell. Highest-impact change: add H1 + lead the intro with the keyword. Everything else is secondary until that's fixed.

Get this skill + 8 more

Get the full SEO & Content pillar (9 skills) or the complete library.

Get the full stack — $299

What you get

310-line SKILL.md, ready to drop into ~/.claude/skills/
Tested through 3 Karpathy-loop iterations (versions v1.0.0 → v1.3.0)
Triggers automatically when relevant — no command to remember
Lifetime updates as the skill is refined further