# Three SEO skills on pre-launch Astro: what each caught > Three SEO skills, one pre-launch Astro site, no live URL: title fixed in 3 minutes, full-site scored at 82/100, content E-E-A-T at 65/100. **Canonical URL**: https://agentcookbooks.com/blog/three-seo-skills-pre-launch-audit/ **Published**: 2026-04-29 **Tags**: claude-code, skills, seo --- Three skills, one site, one day before the custom domain attached. No Search Console, no paid SEO tools, no live URL to crawl — everything ran against Astro source files. The targets: [`/seo-page`](/skills/seo-page/) on `src/pages/index.astro`, [`/seo-audit`](/skills/seo-audit/) on the full source tree, and [`/seo-content`](/skills/seo-content/) on a single skill wiki entry. The receipts: per-page audit caught a 16-character title tag and a missing internal link in 3 minutes; full-site pass scored 82/100 with a 60% framework-degradation footnote (no live URL means no crawl, no Core Web Vitals, no indexation data); content audit landed E-E-A-T at 65/100 and surfaced the placeholder-receipts gap on a bulk-imported wiki entry. Below: what each skill found and where the degradation kicks in without paid tooling. ## [`/seo-page`](/skills/seo-page/): per-page audit, two fixes in 3 minutes **Target:** `src/pages/index.astro` (homepage). The skill loads about 85 lines of audit framework — on-page checklist, content quality, schema, images, Core Web Vitals, a full output-format spec. One guardrail baked in: never recommend HowTo (deprecated) or FAQ (government/health-only) schema. That alone saves a common mistake on pages with Q&A-shaped sections. Two critical fixes landed immediately. **1. Title tag was 16 characters.** The homepage wasn't passing a `title` prop to `Base.astro`; the layout renders `${title} — ${SITE_TITLE}` if a prop is passed, otherwise just `SITE_TITLE`. Result: the page title was just "Agent Cookbooks." Fixed with `title="Claude Code cookbook: hooks, skills, MCP"` — 58 characters, primary keyword front-loaded. **2. No body link to `/skills/`.** The site had 142 skill wiki entries at that point and one blog post. The homepage body linked only to `/blog/`. One-word anchor added. A third critical issue — `/og-default.png` and `/logo-512.png` referenced in code but not generated — wasn't file-editable. Flagged, not fixed. **The source-only catch:** the skill is shaped for live HTML. Applied to an Astro source file it means manually tracing the layout chain — `index.astro` → `Base.astro` → `consts.ts` — to know what actually ends up in ``. Three reads before the audit can start. Not a skill defect; the prescribed path uses DataForSEO MCP on a live URL. Just not what was available. Time: compressed a ~10-minute manual check to ~3 minutes by enumerating exactly what to look at. ## [`/seo-audit`](/skills/seo-audit/): full-site pass, 82/100 **Scope:** 8 source files read in parallel — `robots.txt`, `index.astro`, `skills/index.astro`, `blog/index.astro`, `Base.astro`, `SkillPage.astro`, `BlogPost.astro`, `llms.txt.ts`, `astro.config.mjs`. The skill loads about 340 lines of methodology: Crawlability/Indexation, Technical Foundations, On-Page, Content Quality, E-E-A-T, and a substantial International SEO/hreflang reference section (not applicable here — single-locale site, ~80 lines loaded for nothing). One warning earns its keep immediately: `web_fetch` can't see JS-injected schema. A lesser audit would false-flag missing schema on raw HTML reads. All schema on this site is in `