dobroslavradosavljevic SEO
Complete agent guidance for search, snippets, and AI citations. Classify mode and surface first. Load only the matching references. Do not treat dashboards like landing pages.
Refresh dated claims from source-map.md when behavior looks version-sensitive.
Workflow
- Ground in the project: public origin, trailing-slash and www policy, locales, existing
head/ sitemap / robots, indexable vs app routes. - Pick mode (default: implement). Pick surface and index intent from surfaces.md.
- Load the files in the map below. Do not load the whole folder for a
noindexsettings page. - Decide index intent before writing a title. Then crawl/index → metadata → copy (copywriting.md) → links → schema (only if eligible).
- Emit the decision record (or brief / audit findings). Verify with review.md.
Rules stay stack-agnostic. When the app is TanStack Start / React SSR, prefer route head + loaders + HeadContent and server routes for sitemap.xml / robots.txt — see rendering.md. Otherwise use the project's existing head and sitemap APIs.
Mode map
| Mode | User signal | Load |
|---|---|---|
| Implement | add/fix a public page | surfaces + crawl-index + metadata + copywriting + on-page + (schema/links if needed) |
| Brief | outline / SEO brief | content-research + copywriting + on-page + internal-linking |
| Research | what should we rank for | content-research (+ review measurement if GSC exists) |
| Audit | review / why preview is wrong | review + surfaces + the failing layer |
| IA / links | structure, orphans, cannibalization | information-architecture + internal-linking |
| Local | NAP, maps, locations | local + structured-data |
| i18n | locales, hreflang | i18n + crawl-index |
| GEO | AI Overviews, citations, llms.txt | geo + copywriting + on-page + ee-at |
| Ecommerce | products, facets, variants | ecommerce + crawl-index + structured-data |
| Off-page | links, PR, mentions | off-page only (white-hat; no outreach mail) |
Surface → extra files
| Surface | Default index | Also load |
|---|---|---|
| Marketing (landing, pricing, compare, changelog) | Index | copywriting, on-page, metadata, media, structured-data |
| Content (blog, docs, guides, glossary) | Index | copywriting, on-page, ee-at, internal-linking, structured-data |
| SaaS app (auth, settings, billing, dashboards) | noindex |
metadata, crawl-index |
| Marketplace (listings, products, profiles) | Index unique public entities | copywriting, ecommerce, crawl-index, structured-data |
| Local (locations, contact) | Index real places | local, structured-data, media |
| Legal | Index if they should be found | on-page (thin + honest) |
Rendering / CWV: rendering.md, performance.md.
Head tags: metadata.md, media.md.
Hard rules
- One indexable URL → unique title, unique description, one primary H1, one primary intent.
- Copy is people-first and specific. Title, H1, deck, and CTA agree. No stuffing, throat-clearing, or invented proof. Follow in-repo voice docs when they exist.
- Canonical,
og:url, andog:imageare absolute HTTPS. One host, one slash policy, HTTPS only. robots.txtis notnoindex. Sitemap lists only 200 + canonical + indexable URLs. Honestlastmod.- If the URL must rank or preview, title, primary copy, and
<a href>links exist in the first HTML. - JSON-LD matches visible content. No fake FAQ, ratings, prices, or addresses.
- Faceted/sort/session URLs are not an indexable graph. Paginated pages self-canonicalize; crawlable
hrefs. - Keyword research informs URL and H1. Do not invent volume/KD. Do not ship doorway or scaled city pages.
- E-E-A-T is a quality framework, not a ranking-factor toggle. Do not fabricate authors, credentials, or stats.
- GEO does not replace SEO.
llms.txtis unofficial and optional. Same page must help humans. - Local: one NAP everywhere. Google Business Profile actions stay with the user.
- Off-page: earn, do not buy. Stop at an asset + target list. No PBNs, cloaking, or parasite SEO.
- Do not claim unpublished ranking factors. Core Web Vitals help; they do not guarantee rank.
Decision record (implement / audit)
## URL / route
## Surface + index intent
## Primary query + intent
## Title / H1 / description
## Voice / CTA / claims
## Canonical / robots / sitemap
## Internal links in / out
## Schema (type or none) + visible parity
## GEO note (citeable claim or n/a)
## Verification
Content brief (research / brief)
Use the template in content-research.md. Fill proof and non-goals. No word-count targets.
References
| File | Use |
|---|---|
| surfaces.md | Index defaults; must-include by page type |
| crawl-index.md | robots, canonical, sitemap, status, URL hygiene |
| metadata.md | Title, description, head, OG minimum |
| rendering.md | SSR/SPA, first HTML, Start examples |
| performance.md | LCP, INP, CLS |
| copywriting.md | Voice, answer-first copy, CTAs, page-type writing, anti-slop |
| on-page.md | Intent, headings, content types, snippets |
| ee-at.md | Experience, expertise, trust, YMYL |
| information-architecture.md | Topology, nav, cannibalization, facets |
| internal-linking.md | Anchors, hubs, link-plan template |
| content-research.md | Keywords, SERP, briefs, next 5 pages |
| structured-data.md | JSON-LD types, visible parity, validation |
| geo.md | AI Overviews, citeable pages, llms.txt, bots |
| i18n.md | hreflang, locales |
| local.md | NAP, GBP vs code, LocalBusiness |
| ecommerce.md | Products, variants, facets |
| media.md | OG images, image/video SEO |
| off-page.md | White-hat links and PR |
| review.md | Audit, severity, GSC measurement |
| source-map.md | Official URLs to refresh |
Isolation
This skill stands alone. Reference only the current repo, official docs, and this skill's references/.
Out of scope
Paid ads, black-hat, PBNs, cloaking, automated outreach email, Search Console API productization, invented traffic/KD numbers, and visual design systems.