agentsclimarketplace
Sign in

Seo hreflang

Skill amirjahfar1/automate-seo-with-claude/skills/seo-hreflang

Hreflang and international SEO audit for multi-language and multi-region sites. Validates language-region codes, return tags, x-default, canonical alignment, and conflict detection across the per-URL HTML (DataForSEO On-Page + Firecrawl/WebFetch) and the XML sitemap. Use when the user asks "hreflang", "international SEO", "i18n", "language targeting", "x-default", "regional sites", or "multi-language SEO".From its SKILL.md

Install
npx -y skills add amirjahfar1/automate-seo-with-claude --skill seo-hreflang

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

3 things to look at

  • 2 stars2 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.
  • runs commandsInstructs the agent to run 1 command, including `python3 scripts/gsc_query.py --property "{property}" --json`.
  • fetches URLsInstructs the agent to fetch 2 URLs, including https://{domain}/sitemap.xml and 1 more.

SKILL.md

15.8 KB, ~4.0k tokens by cl100k_base, as published. Nobody here has run it

Example output: examples/seo-hreflang-airbnb-com-20260514/HREFLANG-REPORT.md

Hreflang Audit

Adapted from AgriciDaniel/claude-seo's seo-hreflang skill (MIT). Concept and validation rules originate there; this implementation is rebuilt against our backend (DataForSEO On-Page loop + Firecrawl + WebFetch + GSC + Google APIs via seo-google).

Validate hreflang implementations on a multi-language or multi-region site. Read the actual <link rel="alternate" hreflang="…"> tags pages emit — primarily from the DataForSEO on_page_instant_pages loop (which returns each page's hreflang set) plus Firecrawl/WebFetch for raw <head> markup — cross-check the XML sitemap's hreflang, weight by which country/language pages get impressions in GSC, and produce one of three verdicts — PASS, NEEDS-FIX, or BROKEN — with a top-fixes table anchored in objective signals.

Prerequisites

  • DataForSEO MCP server connected — on_page_instant_pages returns each fetched page's hreflang set, the primary signal source here.
  • GSC (mcp__gscServer__*) recommended — for which country/language pages actually get impressions (traffic-weighting the audit) and cross-domain hreflang verification (step 6). Firecrawl optional — for raw <head> HTML where exact <link rel="alternate"> attributes matter; WebFetch is the fallback.
  • Claude's WebFetch tool available (fallback when Firecrawl is unavailable).
  • User provides: a target domain (e.g. example.com). Optional: explicit list of representative pages to inventory; explicit sitemap URL if not at /sitemap.xml.
  • Predecessor (recommended): seo-technical-audit or seo-sitemap already run on this domain — its discovered URL set + On-Page hreflang fields can be reused. Without it, the skill discovers and fetches its own sample.

Optional accelerator: if you have a hosted-crawl MCP you can substitute it for the URL-discovery + fetch loop — not required.

Process

  1. Validate target & preflight. See skills/seo-firecrawl/references/preflight.md for the canonical 3-stage preflight (cost note, Firecrawl availability, Google APIs). Skill-specific notes:

    • Normalise domain (strip protocol, trailing slash) before continuing.
    • DataForSEO bills per call; this run issues ~5–10 calls (the homepage + ~5-sample on_page_instant_pages loop, plus up to ~6 Firecrawl credits for raw <head> markup). Use the documented limit/ceiling params to cap.
    • Firecrawl: optional with WebFetch fallback, ~6 Firecrawl credits if available (hard cap). When available, step 4 (per-URL hreflang inventory) runs on homepage + 5 representative pages with formats: ["rawHtml"]. Without Firecrawl, step 4 falls back to WebFetch — coverage is degraded because WebFetch returns markdown only and silently strips <link rel="alternate"> tags from <head>. Pass --no-firecrawl to force WebFetch even when Firecrawl is available.
    • Google APIs: tier 1 (GSC) unlocks step 6 (GSC verification of hreflang-targeted alternates). See skills/seo-google/references/cross-skill-integration.md for the full enrichment contract.

  2. Discover the URL set mcp__firecrawl-mcp__firecrawl_map / WebFetch (sitemap) + mcp__gscServer__get_search_analytics + mcp__dataforseo__dataforseo_labs_google_relevant_pages

    • Build the candidate URL set for the audit: XML sitemap entries (especially any language-path branches like /en/, /fr/, /de/), GSC indexed pages (dimensions=["page"] — surfaces which country/language URLs Google actually serves), and DataForSEO top pages.
    • If seo-technical-audit or seo-sitemap already ran, reuse their discovered URL set instead of re-discovering.
    • Cap to a representative sample (homepage + ~5 pages spanning the language paths). State the sample in the deliverable.

  3. Pull On-Page hreflang fields mcp__dataforseo__on_page_instant_pages (looped)

    • Loop on_page_instant_pages over the sampled URLs — each result includes the page's <link rel="alternate" hreflang> set, its canonical, and status. This is the primary hreflang signal source (replacing a hosted crawl's hreflang report).
    • Extract every hreflang issue from these fields (typical codes: hreflang_no_return_tag, hreflang_invalid_lang_code, hreflang_conflict, hreflang_missing_x_default, hreflang_canonical_mismatch, hreflang_no_self_reference).
    • Where the On-Page fetch's hreflang set looks incomplete (some sites inject alternates client-side or in unusual head order), confirm against the raw <head> markup in step 4.
    • Persist to 01-audit-hreflang-issues.md and feed into hreflang-issues.csv.

  4. Per-URL hreflang tag inventory mcp__firecrawl-mcp__firecrawl_scrape (preferred) / WebFetch (fallback)

    • Sample selection: homepage + up to 5 representative pages from mcp__dataforseo__dataforseo_labs_google_relevant_pages (sort by traffic descending; bias toward pages on different language paths if the URL structure exposes them — /en/, /fr/, /de/, etc.).
    • Firecrawl path (1 credit per URL, ~6 total): call firecrawl_scrape(url=..., formats=["rawHtml"]). Pin rawHtml — the default html post-processing strips <link rel="alternate"> on many sites. Parse every <link rel="alternate" hreflang="…" href="…"> from the <head>. Capture: source URL, hreflang attribute, href, and whether it's self-referencing.
    • WebFetch fallback (no Firecrawl): try fetching each URL and extracting hreflang from the markdown response. WebFetch frequently returns markdown that has stripped <head> link tags, so this path will under-report. Note in HREFLANG-REPORT.md: Per-URL inventory: degraded coverage — Firecrawl not installed; some hreflang tags may be missed.
    • Apply validation rules (see references/validation-rules.md for the full list):
      • Self-referencing tag: the page's own URL must appear in its own hreflang set.
      • Return tags: every alternate link must reciprocate. If page A lists B as fr, page B must list A as en (or whichever).
      • x-default: at least one alternate per set must use hreflang="x-default".
      • Language-region code validation: every value must be a valid ISO 639-1 language (optionally followed by - and an ISO 3166-1 Alpha-2 region). Common errors caught: eng (use en), jp (use ja), en-uk (use en-GB), es-LA (no such ISO region).
      • Conflict detection: the same hreflang value (e.g. de-DE) appearing on multiple distinct URLs is a conflict — Google ignores conflicting sets.
      • Canonical alignment: if the page has <link rel="canonical">, it must match the page's own URL (or its self-referencing hreflang URL). Hreflang on a non-canonical page is silently ignored by Google.
      • Protocol consistency: all URLs in a set must share the same scheme (HTTPS preferred).
    • Persist to 02-per-url-hreflang.md and append findings to hreflang-issues.csv.

  5. Sitemap-level hreflang (defer to seo-sitemap where appropriate)

    • If the user's domain uses sitemap-based hreflang (<xhtml:link rel="alternate" …> inside the sitemap), this skill checks structure and consistency only. Full sitemap analysis (orphans, missing pages, broken entries) is seo-sitemap's job — recommend it explicitly if a sitemap-vs-audit diff is in scope.
    • Fetch the sitemap. Try https://{domain}/sitemap.xml; if 404, fetch /robots.txt and find Sitemap: directives. For sitemap-of-sitemaps, recursively fetch each child.
    • Validate hreflang within the sitemap:
      • Does the sitemap use the xmlns:xhtml="http://www.w3.org/1999/xhtml" namespace? Required for hreflang in sitemaps.
      • Does each <url> entry that has hreflang alternates include itself in the alternate set (self-reference)?
      • Does every alternate listed in one <url> entry reciprocate as its own <url> entry with the same alternate set (return tags)?
      • Are language-region codes valid (apply same rules as step 4)?
    • Cross-check against per-URL inventory (step 4): if a sample URL's HTML lists 4 hreflang alternates but the sitemap entry for that URL lists 6, that mismatch is a conflict — Google may pick either, and inconsistency degrades the signal.
    • Persist to 03-sitemap-hreflang.md and append findings to hreflang-issues.csv.

  6. GSC verification of hreflang-targeted alternates (only if google-api.json is present, tier ≥ 1)

    • For each unique domain that appears as an href target in the hreflang sets (e.g. example.com, example.de, example.fr), confirm GSC verification: python3 scripts/gsc_query.py --property "{property}" --json (a status-only check; just confirm the property responds without PROPERTY_NOT_VERIFIED).
    • Why this matters: Google explicitly recommends verifying every domain that participates in a cross-domain hreflang setup. If example.de is listed as an alternate but isn't verified in this account, the hreflang signal is weakened and you can't see how Google interprets it.
    • Surface in HREFLANG-REPORT.md as a section "## GSC verification of hreflang targets" with one row per target domain: verified / not verified / not configured.
    • If property not verified for a target domain: list it as a fix at Medium severity ("Verify {domain} in Google Search Console — required for cross-domain hreflang trust").
    • See skills/seo-google/references/cross-skill-integration.md § "Trigger pattern" for the failure-mode contract.

  7. Synthesise verdict

    • Apply the verdict heuristic (see Tips) to produce PASS, NEEDS-FIX, or BROKEN.
    • Sort hreflang-issues.csv by severity descending, then count descending.
    • Write HREFLANG-REPORT.md with the top-fixes table (top 10) and the verdict.

Output format

Create a folder seo-hreflang-{target-slug}-{YYYYMMDD}/ with:

seo-hreflang-{target-slug}-{YYYYMMDD}/
├── 01-audit-hreflang-issues.md   (On-Page hreflang-field findings filtered to hreflang)
├── 02-per-url-hreflang.md         (per-URL <link rel="alternate"> inventory + validation findings)
├── 03-sitemap-hreflang.md         (sitemap-level hreflang validation; defer details to seo-sitemap)
├── evidence/
│   ├── homepage-rawhtml.html      (raw HTML from Firecrawl, for the homepage sample)
│   ├── sample-{n}-rawhtml.html    (raw HTML for each sampled URL)
│   └── sitemap.xml                (raw fetched sitemap)
├── hreflang-issues.csv            (load-bearing: URL, issue code, severity, fix)
└── HREFLANG-REPORT.md             (PRIMARY: verdict + top fixes table)

HREFLANG-REPORT.md follows this shape:

# Hreflang Audit: {domain}

> Audit date {YYYY-MM-DD} · Sample size: {n} URLs · Languages detected: {comma-separated list}

## Verdict: {PASS | NEEDS-FIX | BROKEN}

Reasoning: {1–2 sentences anchored in concrete numbers from the data}.

## Summary

| Source | Findings | Severity breakdown |
|---|---|---|
| On-Page hreflang fields | {n} | Critical: {n} · High: {n} · Medium: {n} · Low: {n} |
| Per-URL HTML inventory | {n} | … |
| Sitemap | {n} | … |

## Top fixes (impact-ranked)

| # | URL | Issue | Severity | Fix |
|---|---|---|---|---|
| 1 | {URL} | {issue code} | {severity} | {one-line fix} |
| 2 | … | … | … | … |
| ... up to 10 |

## Languages detected

| Language | URL count | Self-ref OK | Return tags OK | x-default OK |
|---|---|---|---|---|
| en-US | {n} | ✓ / ✗ {count} | ✓ / ✗ | ✓ / ✗ |
| de-DE | {n} | … | … | … |
| ... |

## Per-URL inventory ({n} URLs sampled)

| URL | Alternates | Self-ref | x-default | Notable issues |
|---|---|---|---|---|
| {URL} | {n} | ✓/✗ | ✓/✗ | {short text} |
| ... |

## Sitemap-level hreflang
- xhtml namespace declared: {✓/✗}
- URLs with hreflang alternates: {n}
- Self-reference within sitemap: {✓ all / ✗ {count} missing}
- Return tags within sitemap: {✓ all / ✗ {count} missing}
- Per-URL HTML vs sitemap consistency: {✓ all match / ✗ {count} mismatched}
- Full sitemap-vs-audit analysis: see `seo-sitemap` (orphans, broken entries, lastmod).

## GSC verification of hreflang targets

| Domain | Verified | Notes |
|---|---|---|
| {domain1} | ✓ / ✗ | {note if unverified} |
| ... |

(Or: `GSC verification: not configured (run bash extensions/google/install.sh)`.)

## Coverage notes

- Per-URL inventory tool: {Firecrawl rawHtml | WebFetch fallback (degraded — some hreflang tags may be missed)}.
- Pages sampled: homepage + {n} representative pages (selection: top traffic from `dataforseo_labs_google_relevant_pages`).

## Apply

- Walk `hreflang-issues.csv` row-by-row; each row is one specific change (URL + issue + fix).
- After applying changes, re-run `seo-technical-audit` to refresh the On-Page crawl baseline, then re-run this skill to verify.

hreflang-issues.csv columns: url,issue_code,severity,fix,source where source is one of audit | html | sitemap | gsc (audit = the On-Page hreflang-field pass, step 3).

Tips

  • DataForSEO allows up to 2,000 calls/min, 30 concurrent; pace sequentially.
  • Keep the sample small — homepage + ~5 language-spanning pages. The per-URL on_page_instant_pages loop and Firecrawl raw-HTML pulls are the per-call cost here; a representative sample catches systemic hreflang errors without fetching every locale.
  • Verdict heuristic:
    • PASS: zero Critical findings; ≤ 2 High findings; sample URLs all have self-reference, x-default, and reciprocal return tags; all language-region codes valid; canonical aligns with self-ref hreflang.
    • NEEDS-FIX: any High finding; or > 5 Medium findings; or any one of (missing x-default, missing return tags on > 25% of sampled pages, language-region code error, sitemap-vs-HTML mismatch).
    • BROKEN: any Critical finding; or hreflang attempted but no self-reference on the homepage; or canonical pointing elsewhere on a page that nonetheless emits hreflang (entire set is ignored by Google); or > 50% of sampled URLs missing return tags.
  • Anchor every claim in HREFLANG-REPORT.md to a row in hreflang-issues.csv. If a stakeholder questions the verdict, walk them through the CSV.
  • For sites with > 50 language variants per page, the per-URL HTML implementation bloats the <head> — recommend the sitemap-based implementation instead. Don't generate code; the deliverable is diagnostic, not code-gen.
  • The skill does not assess cultural adaptation, content parity, or locale formatting. Those are translation/QA concerns; they're orthogonal to whether hreflang itself is technically correct. If the user wants those, point them at the translation team — this skill answers "is the technical hreflang signal working?", not "is the localised content good?".
  • For cross-domain hreflang (e.g. example.com ↔ example.de), step 6's GSC check is the highest-value enrichment — verifying both domains is Google's explicit recommendation.
  • Common false-positive guard: if a sampled page legitimately has no internationalization (e.g. a single-region site), zero hreflang tags is correct, not an issue. The skill detects this by checking whether any sampled page emits hreflang; if none do, the verdict is PASS with note "No hreflang implementation detected — single-language site."
  • Pair with seo-sitemap for the sitemap-vs-audit diff; pair with seo-technical-audit for full technical health. This skill is narrow: hreflang correctness only.
  • See references/validation-rules.md for the full per-issue rule table (severity, detection logic, suggested fix).

What ships with it: 1 file

6.3 KB alongside SKILL.md

references/

Gives 0 of the 12 instructions most marketing audience skills give in ~4.0k 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

  • normalise the domain before processing
  • discover target urls from sitemap and analytics
  • reuse existing url sets if available
  • extract hreflang issues from on-page data
  • inventory hreflang tags from raw html
  • request raw html format when scraping urls

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.