seo-demand-validator · git:20260918.2429d96 · 2026-09-18 · sha256 62496e57a4a80492
seo-demand-validator git:20260918.2429d96A
Immutable. This exact content never changes and is served at /api/v1/blob/62496e57a4a80492.
--- name: seo-demand-validator sheetId: "1.20" description: "Sixth stage of the SEO pipeline: validate the real demand and competitive difficulty of keywords, compare the common patterns across the current SERP, prune unnecessary terms, and consolidate a list of target keywords and topics. Corresponds to node (6) in the v2 flow diagram. Triggers when the user hands over a Classified handoff packet or says 「驗證需求」 (validate demand) or 「整理主題清單」 (organize the topic list). It only validates and converges — it does not write the Brief and does not write the article." category: brand-marketing project: seo-demand-validator platform: claude status: WIP author: Peter Tu input: "seo-keyword-classifier's SEO_PACKET ▸ Classified; manual Google/SERP observations and tool-based difficulty data" process: "Compare SERP common patterns and competitive difficulty per the competitor-review-table → manually validate demand, prune unnecessary terms → converge into a topic list + target keywords → hand off" output: "Topic list + target keywords + SERP common patterns, handed off to seo-brief-writer" synergy: ["seo-brief-writer"] handoff: ["seo-brief-writer"] house-style: bound --- # SEO Demand Validator ```bash npx skills add https://github.com/peter-tu-zynkr/zynkr-skill-builder --skill seo-demand-validator ``` Sixth stage of the SEO pipeline, corresponding to the flow diagram's "search the keywords, manually validate demand" and "assist in validating relevance and competitive difficulty / compare the common patterns across the current SERP / organize the classification and target keywords." It converges the clustered keywords into a topic list that is "worth doing" — validated with real SERP observation, not just tool numbers. --- ## Resources you'll use > **Knowledge source**: The rubrics/templates used by this skill should be read first from the SEO Knowledge Base's "01 Rubrics & Templates" (Drive, google-workspace MCP, search by name); when unavailable, fall back to the local `./references/`. See the mapping table in `seo-article-pipeline/seo-pipeline-config.md`. - **Competitor review table**: `./references/competitor-review-table.md` - **SEO Knowledge Base folder ID**: `<your-seo-kb-folder-id>` - **MCP / tools**: `google-workspace`; WebFetch / WebSearch to view the current SERP ## Step 1 — Receive the clustering Read `SEO_PACKET ▸ Classified`. ## Step 2 — Manually validate demand (HITL) Corresponds to "Google-search each variant to validate demand." Have the user (or use WebSearch) look at the actual SERP for each candidate target term: whether there is real content competing, and what the searcher actually wants. ## Step 3 — Compare SERP common patterns + difficulty Fill in the competitor review table per `./references/competitor-review-table.md`: the common patterns of the top results (format, depth, angle), competitive difficulty, and whether Zynkr can win. Corresponds to the two AI nodes of (6). ## Step 4 — Prune terms + converge the topic list (HITL) Corresponds to "remove unnecessary terms." Cut the terms with weak demand or difficulty too high to have a winning chance, and converge the rest into a "topic → target keywords (primary + supporting)" list. ## Step 5 — Hand off and save Save the "topic list" working file, and output: ``` SEO_PACKET ▸ Topics - 主題:<標題> |目標關鍵字:<主字 + 輔字> |意圖/漏斗:<...> |SERP 共通點:<...> |Zynkr 勝出點:<...> - ... 主題清單與目標關鍵字已確認,可交棒給 seo-brief-writer。 ``` ## Outputs Topic list + target keywords + SERP common patterns (`SEO_PACKET ▸ Topics` + Drive working file). ## Limitations Does not write the Brief and does not write the article. Difficulty data is grounded in tools / the real SERP, not guessed. ## House style Writing style is **not owned by this file**. The house voice lives in two Google Docs under `[@] 寫作指南` (`12DBdFz3SK22ie9im_ThFMI7IBRXsTZsV`), read at runtime: - 《[2.0] Zynkr 通用風格指南 House Voice》 `10bOIQwRm9Pxwgct4hlwCwK_B4Pipai1HqBPZKzyRHSE` — the universal core, plus the addendum for this surface - 《[3.2] 禁用詞清單 Forbidden Words》 `1N5sHLP4qzmmhpCGsi6KElxi1z0MFe4QZ0Q_35T10Uyg` Read both before producing client- or reader-facing text, and scan the draft against 《[3.2]》 before handing it over. If Drive is unreachable, say so in the output rather than proceeding unchecked. Never re-implement either list inside this file.