agentsclimarketplace

Seo headings structure

Skill Hainrixz/claude-seo-ai/skills/seo-headings-structure

Audit and propose fixes for a page's heading hierarchy and semantic outline — verify exactly one H1, logical H2-H6 nesting with no skipped levels, descriptive heading text, and HTML5 landmarks (article/section/nav/main/footer) — so the document outline is extractable. Module M7c. Feeds the Search SEO score (clean headings also support AI answer-block extraction, which is scored separately under M11).From its SKILL.md

Install
npx -y skills add Hainrixz/claude-seo-ai --skill seo-headings-structure

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

  • runs commandsInstructs the agent to run 2 commands, including `node ${CLAUDE_SKILL_DIR}/../../scripts/parse-html.mjs --url <u>` and 1 more.

SKILL.md

4.7 KB, ~1.1k tokens by cl100k_base, as published. Nobody here has run it

seo-headings-structure (M7c)

A clean, single-rooted heading tree and proper semantic landmarks make a page's outline machine-readable — the same structure that AI engines use to extract answer blocks and that search engines use to understand content scope. Schema/entity context: references/schema-tier1.md.

Audits

Working from the PageSnapshot (rendered_dom if present, else raw_html):

  1. Single H1: exactly one <h1> per page. Zero H1s, or multiple H1s, are both findings.
  2. No skipped levels: nesting descends one level at a time (an <h2> may be followed by <h3>, not directly by <h4>). Build the outline and flag any jump.
  3. Descriptive text: each heading names its section in human terms — flag empty headings, headings used purely for visual styling, and generic placeholders ("Section 1", "Welcome").
  4. Semantic landmarks: detect <main>, <article>, <section>, <nav>, <header>, <footer>; flag when headings live inside non-semantic <div> soup or when more than one <main> exists.
  5. Outline ↔ landmark agreement: sectioning elements should carry their own heading; orphan landmarks and headings outside any landmark are findings.

Good heading structure feeds AI answer-block extraction: clear H2/H3 boundaries map to candidate quotable passages.

Fixes

  • Heading-level restructuring — re-leveling H2-H6 to remove skips or to nest correctly (fixable: proposed). This alters the visual hierarchy, so it requires per-item accept; emit a fix_preview diff, never an auto-write.
  • Add a single missing H1 — when the page has no H1, propose one sourced from the populated <title>/og:title or the obvious page-title element (fixable: proposed, per-item accept — adding a visible heading is editorial). Never fabricate H1 text from thin content; ask the user if the source is ambiguous.
  • Wrap in semantic landmarks — recommending <main>/<article> adoption is advisory (structural refactor the tool does not write).

Mark each finding's fixable field consistent with schema/finding.schema.json (auto / proposed / advisory). Never fabricate a heading's text or invent an H1 from thin content — ask the user or leave a clearly-marked TODO placeholder.

Verification

  • Method: dom_assert. Re-parse the snapshot, rebuild the outline, and assert the specific condition (one H1, no level skip, landmark present).
  • Offline reproduce: node ${CLAUDE_SKILL_DIR}/../../scripts/parse-html.mjs --url <u> (or --file <path>) — returns the headings block (counts per level, h1_count, and skipped_levels), which counts H1s and walks the H1-H6 tree for skipped levels.
  • When the rendered DOM tier is required but unavailable (heading injected client-side and only raw_html is present), status is needs_api, never a false pass.

Findings

Emit findings per schema/finding.schema.json. Examples:

  • M7c.h1.multiple — more than one <h1> on the page (status fail, severity 3, fixable: proposed, axis search, confidence established). evidence.observed quotes each H1's text and selector; verification.reproduce is the parse-html.mjs command above (inspect h1_count).
  • M7c.outline.skipped_level — an <h2> followed directly by an <h4> (status fail, severity 3, fixable: proposed, axis search, confidence directional). evidence.observed quotes the two adjacent headings and their levels.
  • M7c.landmark.no_main — no <main> element; primary content sits in <div>s (status warn, severity 3, fixable: advisory, axis search, confidence directional).

Each finding carries evidence.observed quoting the page verbatim and a runnable verification.reproduce; expected_impact is banded + confidence-tagged (no naked percentages).

Honesty

  • Heading keyword stuffing does not buy ranking — write headings for the reader; clarity is the signal, not keyword density. Don't recommend cramming terms into H1/H2.
  • Multiple H1s are valid HTML5 and modern parsers/AI tolerate them; the real cost is a muddied outline, so treat as severity 3 directional, not a critical failure. Skipped heading levels are primarily an accessibility/structure concern — frame impact honestly (directional) rather than promising a ranking lift.

What ships with it

Read from the repository

Just SKILL.md. No reference files, no scripts.

Gives 0 of the 12 instructions most marketing audience skills give in ~1.1k tokens

Counted across 690 of the 894 authors here whose files we hold, read 2026-08-07

  • Apply Poppins font to headingsin 41 of 690, across 6 files
  • Apply Lora font to body textin 41 of 690, across 6 files
  • Use Arial fallback for headingsin 39 of 690, across 4 files
  • Use Georgia fallback for body textin 39 of 690, across 4 files
  • Maintain text hierarchy and formattingin 39 of 690, across 4 files
  • Use accent colors for non-text shapesin 38 of 690, across 3 files
  • Use RGB values for precise color matchingin 38 of 690, across 3 files
  • Use brand colors for primary text and backgroundsin 36 of 690, across 1 file
  • Read product marketing context file before asking questions, starting, or auditingin 35 of 690, across 23 files
  • Use active voice instead of passive voicein 26 of 690, across 10 files
  • Implement or generate appropriate JSON-LD structured datain 24 of 690, across 17 files
  • Prioritize clarity over clevernessin 22 of 690, across 8 files

Said here and by no other author read

  • quote observed page content as evidence
  • flag any skipped heading levels
  • flag empty or generic placeholder headings
  • detect semantic landmark elements
  • flag headings outside any landmark
  • propose heading level restructuring as a diff

Grouped from the skills themselves: near-identical wordings counted once, and counted by distinct author, so one author publishing three of these counts once. Length counted with cl100k_base; the agent that loads this file may tokenize it differently.

Keep looking

Skills are one crate of 325,949. Ordering is by how many stacks a row turns up in, so the top of any crate is what has actually been picked rather than what has the most stars.