` length (the real string, not markdown's first heading), meta description length, `og:title`, `og:description`, `og:image`, `twitter:card`. - From the returned `html`: every `<script type="application/ld+json">` block. Parse and list `@type`s per winner (Article, FAQPage, BreadcrumbList, Product, etc.) — these become the "schema baseline" for the new article. - On-page signals: hero-image presence, byline structure (`<a rel="author">`, `<meta name="author">`), table count, code-block count. - **Classify the brief's template** against the 8-template map in `references/intent-template-map.md`. Cross-reference: (a) the dominant page type across the SERP top-10 (use the heuristics in `skills/seo-sxo/references/page-type-patterns.md`), (b) PAA patterns from step 5, (c) the keyword's intent classification from `DATA_getRelatedKeywords`. Pick one of: `ultimate-guide` / `how-to` / `listicle` / `explainer` / `comparison` / `review` / `best-of` / `landing-page`. If the SERP is split across types, follow the MIXED rule (see Tips). The chosen template determines the recommended H1/H2 outline shape and word-count floor — record both the template and a one-sentence justification in `BRIEF.md`. - **If Firecrawl unavailable (or `--no-firecrawl` passed):** WebFetch portion runs unchanged. The brief's "Top 3 winners — on-page benchmark" subsection (see Output) emits `(skipped — Firecrawl required for schema/og:* on competitor pages)`. Template classification still runs from WebFetch + SERP data. 7. **Internal linking plan** - `DATA_getDomainKeywords` filtered to the target domain plus WebFetch of 5 high-ranking pages on topically adjacent queries. - For each: propose an anchor text and the section of the new post it belongs in. 8. **Synthesise the brief** (see Output Format). ## Output format Create a folder `seo-content-brief-{target-slug}-{YYYYMMDD}/` with the synthesised brief at the top level and step files in `evidence/`: ``` seo-content-brief-{target-slug}-{YYYYMMDD}/ ├── BRIEF.md (writer-ready synthesis — primary deliverable; inlines 01-domain-overview, 02-competitors, 06-internal-links into a "Context" section) └── evidence/ ├── 01-domain-overview.md (DATA_getDomainOverviewWorldwide raw — preserved for reproducibility) ├── 02-competitors.md (DATA_getDomainCompetitors raw) ├── 03-keyword-gaps.md (DATA_getDomainKeywordsComparison filtered) ├── 04-serp-and-keywords.md (DATA_getSerpResults + related/question keywords) ├── 05-content-analysis.md (top-3 winners' H-spine + on-page benchmark) └── 06-internal-links.md (target domain pages + proposed anchors) ``` Top-level: `BRIEF.md` only — a freelance writer should not need to open anything else. Step files preserve the raw API/scrape outputs in `evidence/` for reproducibility / to back up the editor brief. `BRIEF.md` follows this shape: ```markdown # Content Brief: {proposed title} **Template type:** {one of: ultimate-guide / how-to / listicle / explainer / comparison / review / best-of / landing-page / MIXED} **Why this template:** {one-sentence justification grounded in SERP top-10 page-type majority + PAA pattern + keyword intent} ## Target keyword - Primary: {kw} ({volume}/mo, KD {kd}, intent: informational) - Secondary: {kw1} ({volume}), {kw2} ({volume}), {kw3} ({volume}) ## Title options 1. {title option 1} 2. {title option 2} 3. {title option 3} ## Meta description (150-160 chars) {draft} ## Suggested structure ### H1: {proposed H1} #### H2: {section 1} Cover: {bullets of what to include} Cite: {sources to link out to} #### H2: {section 2} ... ## Gaps the current top 3 miss - {gap 1, with evidence} - {gap 2, with evidence} - {gap 3, with evidence} ## Top 3 winners — on-page benchmark (Firecrawl) | Signal | Winner 1 | Winner 2 | Winner 3 | Required for parity | |---|---|---|---|---| | `<title>` length (chars) | {n} | {n} | {n} | {target} | | Meta description length | {n} | {n} | {n} | {target} | | `og:image` present | {✓/✗} | {✓/✗} | {✓/✗} | {yes/no} | | `twitter:card` | {value} | {value} | {value} | {value} | | JSON-LD types | {Article, …} | {…} | {…} | {must include} | | Byline structure (DOM) | {✓/✗} | {✓/✗} | {✓/✗} | {yes/no} | | Word count (rendered) | {n} | {n} | {n} | {target range} | (Or: `Top-3 on-page benchmark: skipped — Firecrawl not installed.`) ## Internal linking plan | From existing page | Anchor text | Target section | |---|---|---| | {url} | {anchor} | {section} | ## AI Search angle - LLMs currently cite {brands} for this query. - To earn mentions: {specific actions, e.g., add a comparison table, cite a primary study, include a structured data block}. ## Deliverables - Word count target: {n} - Tone and voice: {guidance from domain overview} - Required assets: {images, schema, examples} ## Traffic potential - Conservative: ~{n}/mo at position 5 - Target: ~{n}/mo at position 1-3 ## Raw data references - evidence/01-domain-overview.md - evidence/02-competitors.md - evidence/03-keyword-gaps.md - evidence/04-serp-and-keywords.md - evidence/05-content-analysis.md - evidence/06-internal-links.md ``` ## Tips - Respect SE Ranking Data API rate limit: 10 requests per second. Iterate sequentially, do not fan out across 20 keywords in parallel. - **Firecrawl cost.** Step 6's competitor benchmark adds 3 Firecrawl credits per run (1 per top-3 winner). Pass `--no-firecrawl` to skip it (the brief still ships, just without the on-page benchmark table). - If the user only provides a domain and no topic, run step 3 first and present the top 3 gap opportunities before deciding. - Keep the brief self-contained. A freelance writer should not need to open the raw-data files unless they want to double-check something. - Do not hallucinate keyword difficulty or volume. If an endpoint returns null, mark the field unknown in the brief rather than guessing. - If the topic triggers AI Overviews, the AI Search angle section is mandatory, not optional. Growth in this era requires being citable by LLMs. - If the SERP top-10 has mixed page types (e.g. 4 listicles + 4 comparisons + 2 explainers), treat as MIXED and produce a hybrid brief that explains the consultant's chosen template + why; don't force-pick one when the SERP itself is split. Record `MIXED` in the `Template type` line and use the `Why this template` line to justify the hybrid choice (which template's outline shape you're inheriting and which secondary patterns you're folding in).

--- name: seo-content-brief description: Generate a writer-ready SEO content brief from a target domain and topic. Pulls domain overview, competitors, keyword gaps, SERP analysis, related and question keywords, AI Search citations, and existing internal-link sources, then synthesises a complete editorial brief a freelance writer can start from immediately. Use when the user asks for a content brief, blog brief, article outline, editor brief, or wants to capture organic traffic their competitors have. --- > Example output: [examples/seo-content-brief-vercel-rate-limiting-20260514/BRIEF.md](../../examples/seo-content-brief-vercel-rate-limiting-20260514/BRIEF.md) # Content Brief Turn a domain plus a topic intent into a complete content editor brief: target keyword, title options, H2/H3 structure, content gaps the current top results miss, internal linking plan, AI search angle, and estimated traffic potential. ## Prerequisites - SE Ranking MCP server connected. - Claude's `WebFetch` tool available (used for top-3 content teardown). - User has provided: (a) target domain, (b) market/country (default: `us`), and optionally (c) a seed topic or intent. If no seed topic is given, discover the best opportunity from the keyword-gap step. ## Process 1. **Domain overview** `DATA_getDomainOverviewWorldwide`, `DATA_getDomainKeywords` - Pull organic traffic, top countries, and top 100 organic keywords in the target market. - Save the raw JSON and a human summary. 2. **Competitor discovery** `DATA_getDomainCompetitors` - Identify the top 5 organic competitors by shared keywords in the target market. - Save a one-line positioning note per competitor. - **Note:** the upstream API does not support `limit`/`offset`, so this call returns the full set (~60KB for popular domains) and the MCP harness writes it to a file. Read that file path, parse the `{data: [...]}` JSON, sort by `common_keywords` desc, and take the top 5. 3. **Keyword gap analysis** `DATA_getDomainKeywordsComparison` - Pull keywords the competitors rank for that the target domain does not. - Filter: informational intent, search volume > 1,000/mo, keyword difficulty < 40. - Save the filtered gap list sorted by volume. 4. **Pick the best topic cluster** - From the gaps, select one topic. Justify the pick with: traffic potential, difficulty, relevance to the target domain's product. Surface reasoning to the user before proceeding. 5. **SERP and keyword deep-dive for the chosen topic** - `DATA_getSerpResults` for the top 10 organic + SERP features (AIO, PAA, Featured Snippet, Video). - `DATA_getRelatedKeywords` and `DATA_getSimilarKeywords` for expansion. - `DATA_getKeywordQuestions` for People-Also-Ask and question-based variations. - `DATA_getAiOverview` + `DATA_getAiOverviewLeaderboard` to see which brands LLMs cite today for the topic. 6. **Top 3 content analysis** `WebFetch` (always) + `mcp__firecrawl-mcp__firecrawl_scrape` (when available) - **WebFetch first** (free, instant): pull markdown for the top 3 ranking URLs. Extract H1/H2/H3 spine, word count per article, shared subtopics, gaps, and prose-level formatting patterns. - **Firecrawl second** (3 Firecrawl credits — 1 per top-3 winner) — recovers what WebFetch's markdown can't show: - From `metadata`: `` length (the real string, not markdown's first heading), meta description length, `og:title`, `og:description`, `og:image`, `twitter:card`. - From the returned `html`: every `<script type="application/ld+json">` block. Parse and list `@type`s per winner (Article, FAQPage, BreadcrumbList, Product, etc.) — these become the "schema baseline" for the new article. - On-page signals: hero-image presence, byline structure (`<a rel="author">`, `<meta name="author">`), table count, code-block count. - **Classify the brief's template** against the 8-template map in `references/intent-template-map.md`. Cross-reference: (a) the dominant page type across the SERP top-10 (use the heuristics in `skills/seo-sxo/references/page-type-patterns.md`), (b) PAA patterns from step 5, (c) the keyword's intent classification from `DATA_getRelatedKeywords`. Pick one of: `ultimate-guide` / `how-to` / `listicle` / `explainer` / `comparison` / `review` / `best-of` / `landing-page`. If the SERP is split across types, follow the MIXED rule (see Tips). The chosen template determines the recommended H1/H2 outline shape and word-count floor — record both the template and a one-sentence justification in `BRIEF.md`. - **If Firecrawl unavailable (or `--no-firecrawl` passed):** WebFetch portion runs unchanged. The brief's "Top 3 winners — on-page benchmark" subsection (see Output) emits `(skipped — Firecrawl required for schema/og:* on competitor pages)`. Template classification still runs from WebFetch + SERP data. 7. **Internal linking plan** - `DATA_getDomainKeywords` filtered to the target domain plus WebFetch of 5 high-ranking pages on topically adjacent queries. - For each: propose an anchor text and the section of the new post it belongs in. 8. **Synthesise the brief** (see Output Format). ## Output format Create a folder `seo-content-brief-{target-slug}-{YYYYMMDD}/` with the synthesised brief at the top level and step files in `evidence/`: ``` seo-content-brief-{target-slug}-{YYYYMMDD}/ ├── BRIEF.md (writer-ready synthesis — primary deliverable; inlines 01-domain-overview, 02-competitors, 06-internal-links into a "Context" section) └── evidence/ ├── 01-domain-overview.md (DATA_getDomainOverviewWorldwide raw — preserved for reproducibility) ├── 02-competitors.md (DATA_getDomainCompetitors raw) ├── 03-keyword-gaps.md (DATA_getDomainKeywordsComparison filtered) ├── 04-serp-and-keywords.md (DATA_getSerpResults + related/question keywords) ├── 05-content-analysis.md (top-3 winners' H-spine + on-page benchmark) └── 06-internal-links.md (target domain pages + proposed anchors) ``` Top-level: `BRIEF.md` only — a freelance writer should not need to open anything else. Step files preserve the raw API/scrape outputs in `evidence/` for reproducibility / to back up the editor brief. `BRIEF.md` follows this shape: ```markdown # Content Brief: {proposed title} **Template type:** {one of: ultimate-guide / how-to / listicle / explainer / comparison / review / best-of / landing-page / MIXED} **Why this template:** {one-sentence justification grounded in SERP top-10 page-type majority + PAA pattern + keyword intent} ## Target keyword - Primary: {kw} ({volume}/mo, KD {kd}, intent: informational) - Secondary: {kw1} ({volume}), {kw2} ({volume}), {kw3} ({volume}) ## Title options 1. {title option 1} 2. {title option 2} 3. {title option 3} ## Meta description (150-160 chars) {draft} ## Suggested structure ### H1: {proposed H1} #### H2: {section 1} Cover: {bullets of what to include} Cite: {sources to link out to} #### H2: {section 2} ... ## Gaps the current top 3 miss - {gap 1, with evidence} - {gap 2, with evidence} - {gap 3, with evidence} ## Top 3 winners — on-page benchmark (Firecrawl) | Signal | Winner 1 | Winner 2 | Winner 3 | Required for parity | |---|---|---|---|---| | `<title>` length (chars) | {n} | {n} | {n} | {target} | | Meta description length | {n} | {n} | {n} | {target} | | `og:image` present | {✓/✗} | {✓/✗} | {✓/✗} | {yes/no} | | `twitter:card` | {value} | {value} | {value} | {value} | | JSON-LD types | {Article, …} | {…} | {…} | {must include} | | Byline structure (DOM) | {✓/✗} | {✓/✗} | {✓/✗} | {yes/no} | | Word count (rendered) | {n} | {n} | {n} | {target range} | (Or: `Top-3 on-page benchmark: skipped — Firecrawl not installed.`) ## Internal linking plan | From existing page | Anchor text | Target section | |---|---|---| | {url} | {anchor} | {section} | ## AI Search angle - LLMs currently cite {brands} for this query. - To earn mentions: {specific actions, e.g., add a comparison table, cite a primary study, include a structured data block}. ## Deliverables - Word count target: {n} - Tone and voice: {guidance from domain overview} - Required assets: {images, schema, examples} ## Traffic potential - Conservative: ~{n}/mo at position 5 - Target: ~{n}/mo at position 1-3 ## Raw data references - evidence/01-domain-overview.md - evidence/02-competitors.md - evidence/03-keyword-gaps.md - evidence/04-serp-and-keywords.md - evidence/05-content-analysis.md - evidence/06-internal-links.md ``` ## Tips - Respect SE Ranking Data API rate limit: 10 requests per second. Iterate sequentially, do not fan out across 20 keywords in parallel. - **Firecrawl cost.** Step 6's competitor benchmark adds 3 Firecrawl credits per run (1 per top-3 winner). Pass `--no-firecrawl` to skip it (the brief still ships, just without the on-page benchmark table). - If the user only provides a domain and no topic, run step 3 first and present the top 3 gap opportunities before deciding. - Keep the brief self-contained. A freelance writer should not need to open the raw-data files unless they want to double-check something. - Do not hallucinate keyword difficulty or volume. If an endpoint returns null, mark the field unknown in the brief rather than guessing. - If the topic triggers AI Overviews, the AI Search angle section is mandatory, not optional. Growth in this era requires being citable by LLMs. - If the SERP top-10 has mixed page types (e.g. 4 listicles + 4 comparisons + 2 explainers), treat as MIXED and produce a hybrid brief that explains the consultant's chosen template + why; don't force-pick one when the SERP itself is split. Record `MIXED` in the `Template type` line and use the `Why this template` line to justify the hybrid choice (which template's outline shape you're inheriting and which secondary patterns you're folding in).