SEOMatic Keyword Cannibalization

安装指南

1. ClawHub CLI 从源直接安装技能的最快方式。
   npx clawhub@latest install seomatic-cannibalization
2. 手动安装 将技能文件夹复制到以下位置之一
   全局模式
   ~/.openclaw/skills/
   工作区
   /skills/
   优先级：工作区 > 本地 > 内置
3. 提示词安装 将此提示词复制到 OpenClaw 即可自动安装。
   请帮我使用 Clawhub 安装 seomatic-cannibalization。如果尚未安装 Clawhub，请先安装（npm i -g clawhub）。
   更倾向于下载？
   获取包含原生技能文件的ZIP压缩包。
   下载技能 (.zip)

什么是

The SEOMatic Keyword Cannibalization skill uses real Google Search Console data to identify queries where multiple pages on a connected site compete for visibility. It checks each page’s clicks, impressions, and position, filters out harmless overlap such as brand queries, and focuses on conflicts that may be costing clicks or sending visitors to the wrong page.
Rather than making changes automatically, it recommends a winner and a suitable fix—consolidation, retargeting, redirect, or internal links—and can stage tasks in SEOMatic for review. Openclaw Skills users retain control: nothing goes live until the user approves it, and changes can be rolled back.

应用场景

Investigate rankings that fluctuate between two or more URLs for the same query.
Find pages competing in positions 5–30 without a clear winner.
Diagnose when an informational page ranks instead of the intended product or service page.
Identify duplicate-intent content that may be better consolidated.
Decide whether to retarget a page, redirect it, or clarify page relationships with internal links.
Prepare evidence-backed SEO work items for a user to review and approve.

工作原理

Verify that Google Search Console is connected to SEOMatic. If it is not connected, direct the user to https://app.seomatic.ai/connect/gsc and stop; do not substitute estimates for Search Console data.
Call gsc_insights with the cannibalization action, using the default 28-day window and minImpressions of 50 to find candidate queries.
Investigate promising candidates with gsc_performance using query_pages to compare each page’s clicks, impressions, and position. Use dimension_breakdown by date with filterQuery to check whether Google changes the ranking URL over time.
Keep conflicts that show meaningful competition: multiple pages in positions 5–30 without a clear leader, a URL that flips over time, or the wrong page ranking for the query. Exclude brand-query overlap and cases where one page receives over 80% of the clicks.
Crawl the relevant pages with site_pages and select the fix based on search intent, page performance, and page value: consolidate, retarget, redirect, or add internal links.
Present the query, page-level metrics, recommended winner, rationale, and proposed fix. Prioritize a short list of genuine conflicts over a large list of weak signals.
If the user wants work staged and the plan supports it, check existing tasks with task_manage list, then create up to five appropriate tasks. Explain that the tasks await approval; never approve on the user’s behalf.
After the user explicitly approves a specific task, use task_manage decide as appropriate. For redirects, require explicit sign-off on the exact URL and confirmDestructive: true. Verify completed work with task_manage get before reporting that anything changed.

配置指南

Add the SEOMatic MCP server to the Openclaw Skills-compatible agent configuration:
Endpoint: https://app.seomatic.ai/api/mcp
Transport: streamable HTTP
Authentication: OAuth, or a Bearer API key beginning with smk_ from SEOMatic AI agent settings.
Connect the site’s Google Search Console account at SEOMatic GSC connection. Search Console access is required; if it is disconnected, the skill must stop and ask the user to connect it.
Confirm that the MCP server exposes the required tools: gsc_insights, gsc_performance, site_pages, and task_manage.
Use the free tier for the read-only investigation. Staging tasks requires a paid SEOMatic plan and the API key’s agents:act scope. If staging is unavailable, report the recommended fixes without attempting to create tasks.
No CLI installation command or local binary is specified by this skill.
SEOMatic Keyword Cannibalization 数据架构与分类体系
The skill reads site and query performance from Google Search Console through SEOMatic and uses crawled page information to inform recommendations. It does not specify local files or generated artifacts.
Data or record Source/tool Relevant fields or purpose
Cannibalization candidates gsc_insights (cannibalization) Query, competing site pages, 28-day period, minimum 50 impressions by default
Query-page performance gsc_performance (query_pages) Per-page clicks, impressions, and position for a query
Ranking changes over time gsc_performance (dimension_breakdown) Date dimension and filterQuery to identify URL switching
Page content/context site_pages (crawl) Crawl both pages to assess intent and page value
Staged work task_manage Task type, targetKey, user-worded title, evidence-based rationale, and estimated cost label
Supported task types are consolidate, cannib_retarget (requires queries), redirect, and internal_links. targetKey is the full page URL or, for content work, a topic. Redirects are destructive: they require explicit user approval for the specific URL. Staged tasks are not live changes, and no task should be treated as completed until its result is checked.

高级特性

Uses actual connected Google Search Console data rather than guessed rankings or generic duplicate-content checks.
Distinguishes costly cannibalization from normal brand-query overlap and already-resolved cases where one page gets over 80% of clicks.
Detects URL switching over time through date-based performance breakdowns.
Matches recommendations to the situation: consolidate pages with the same intent, retarget pages with different intent, redirect pages without independent value, or use internal links when both pages should remain.
Supports task staging through SEOMatic, with duplicate checks before creation and up to five tasks per creation request.
Includes safeguards for destructive redirects and user-controlled approvals; it never approves work on the user’s behalf.
Allows read-only analysis on the free tier; task staging requires a paid plan with the appropriate
agents:act
scope.
Supports verification by retrieving task results with
task_manage
get
before claiming that a change has taken effect.

name: seomatic-cannibalization description: 'Find keyword cannibalization on the connected site: queries where two or more of its own pages compete in Google, from real Search Console data, decide which page should win, and stage the consolidation, retarget or redirect for approval. Use when the user asks about cannibalization, duplicate content, pages competing, or rankings that flip between URLs. Runs on SEOmatic, the SEO agent that works from the site''s real Google Search Console data; nothing goes live until the user approves it.' version: 1.0.0
license: MIT
homepage: https://app.seomatic.ai