Skip to content
Back to search

qiaomu-seo

SkillWorks with Claude CodeGitHub

Audit, diagnose, research, plan, implement, experiment on, and verify website SEO across Google, Bing, and AI-search surfaces. Use for technical SEO, crawling, rendering and indexing, JavaScript SEO, robots.txt, sitemaps, canonicals, redirects, status codes, metadata, internal linking, structured data, Core Web Vitals, keyword and intent research, content quality or pruning, SEO experiment design, traffic drops, migrations, international SEO and hreflang, ecommerce and product SEO, image or video search, large-site and programmatic SEO, IndexNow, Search Console or Webmaster Tools analysis, and AI Overviews, AI Mode, ChatGPT Search, Copilot, or Perplexity visibility. Use with URLs, website code, rendered pages, server logs, crawl files, first-party exports, and keyword datasets. Exclude paid-search campaign management, app-store optimization, generic LLM prompt optimization, ranking or citation guarantees, link spam, and unsupported metrics or causal claims.

Safety scan

A static text scan, not a guarantee. Select the badge to see what was scanned and every finding.

Checked today
Badge for your README
findsafeskills safety rating

Markdown

[![findsafeskills safety rating](/api/listings/joeseesun--qiaomu-seo/badge.svg)](/listings/joeseesun--qiaomu-seo)

HTML

<a href="/listings/joeseesun--qiaomu-seo"><img src="/api/listings/joeseesun--qiaomu-seo/badge.svg" alt="findsafeskills safety rating" /></a>

Install

i

Install instructions from the README

As written by the repo's author. findsafeskills has not verified these steps.

一行安装到 Codex、Claude Code、Cursor 等兼容 Agent Skills 的工具:

npx skills add joeseesun/qiaomu-seo

安装后可用下面的命令确认 skill 已被发现:

npx skills add joeseesun/qiaomu-seo --list

前置条件

- [ ] 已安装 Node.js:运行 `node --version` 检查;未安装时从 [nodejs.org](https://nodejs.org/) 获取 LTS 版本。 - [ ] 可运行 npm/npx:运行 `npx --version` 检查;它会随 Node.js 一同安装。 - [ ] 使用支持 Agent Skills 的客户端,并允许它读取本机 skills 目录。

What was scanned

The files the safety scan read, as written by the repository's author and shown as plain text. Lines with a finding are marked.

SKILL.md · Manifest · 123 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. ---
  2. name: qiaomu-seo
  3. description: Audit, diagnose, research, plan, implement, experiment on, and verify website SEO across Google, Bing, and AI-search surfaces. Use for technical SEO, crawling, rendering and indexing, JavaScript SEO, robots.txt, sitemaps, canonicals, redirects, status codes, metadata, internal linking, structured data, Core Web Vitals, keyword and intent research, content quality or pruning, SEO experiment design, traffic drops, migrations, international SEO and hreflang, ecommerce and product SEO, image or video search, large-site and programmatic SEO, IndexNow, Search Console or Webmaster Tools analysis, and AI Overviews, AI Mode, ChatGPT Search, Copilot, or Perplexity visibility. Use with URLs, website code, rendered pages, server logs, crawl files, first-party exports, and keyword datasets. Exclude paid-search campaign management, app-store optimization, generic LLM prompt optimization, ranking or citation guarantees, link spam, and unsupported metrics or causal claims.
  4. ---
  5. # Qiaomu SEO
  6. Improve discoverability and qualified organic outcomes with current sources, explicit coverage, and reproducible evidence.
  7. ## Router Rules
  8. - Use for organic website search work: audit, strategy, research, implementation, migration, monitoring, and experimentation.
  9. - Route paid media and bidding to advertising workflows; route app-store listings to ASO; route pure conversion optimization elsewhere unless organic acquisition is also in scope.
  10. - Treat Google, Bing, AI search, image, video, shopping, local, and news as distinct surfaces. Never generalize a feature, crawler, report, or policy across providers.
  11. - Read only the task-relevant modules:
  12. - core audit and prioritization: [Audit Playbook](references/audit-playbook.md)
  13. - crawling, indexing, JavaScript, canonicals, migrations, large sites: [Technical SEO](references/technical-seo.md)
  14. - keywords, intent, page maps: [Keyword And Content](references/keyword-content.md)
  15. - content quality, pruning, programmatic pages: [Content Quality](references/content-quality.md)
  16. - performance, Search Console, incidents, experiments: [Performance And Measurement](references/performance-measurement.md)
  17. - international and ecommerce: [International And Commerce](references/international-commerce.md)
  18. - images and video: [Vertical Search](references/vertical-search.md)
  19. - Google, Bing, IndexNow, OpenAI, Perplexity: [Engine Matrix](references/engine-matrix.md)
  20. - generative search: [AI Search](references/ai-search.md)
  21. - mutable-claim review: [Knowledge Freshness](references/knowledge-freshness.md)
  22. - evidence semantics or JSON output: [Evidence Policy](references/evidence-policy.md) and [Audit Contract](references/audit-contract.md)
  23. ## Action Boundary
  24. - `audit / diagnose / compare / advise`: inspect and report; do not edit code, content, webmaster settings, feeds, DNS, or production.
  25. - `optimize / fix / implement`: change only files or systems the user placed in scope; capture before evidence and verify afterward.
  26. - Never submit URLs, change index controls, publish, delete pages, disavow links, alter Merchant Center or Business Profile, buy data, or contact third parties without explicit authorization.
  27. ## Work Modes
  28. - `advisory`: no target evidence; give a plan without site-specific claims.
  29. - `page`: inspect one or a small named URL set deeply.
  30. - `template sample`: inspect representative templates and disclose selection; never call it a full-site audit.
  31. - `site inventory`: use crawl, sitemap, route, log, or first-party inventories to measure breadth.
  32. - `incident`: diagnose a traffic/indexing loss using segmented timelines and competing hypotheses.
  33. - `migration`: protect URL mappings, redirects, canonicals, hreflang, sitemaps, feeds, monitoring, and rollback.
  34. - `experiment`: define hypothesis, treatment unit, comparison, guardrails, observation window, and decision rule.
  35. - `specialty`: evaluate international, ecommerce, image, video, local, news, or AI-search requirements only when relevant.
  36. ## Compact Workflow
  37. 1. Define outcome, conversion, audience, market/language, search surface, engine/provider, target pages, time window, and authorized action.
  38. 2. Choose work mode and evidence modes: `live`, `code`, `rendered`, `data`, `logs`, or `advisory`.
  39. 3. Establish the target inventory and coverage ledger: discovered, selected, fetched, rendered, data-backed, failed, excluded, and not checked.
  40. 4. Classify required knowledge before using it:
  41. - `stable principle`: durable mechanism such as crawl → render → index.
  42. - `current platform rule`: provider documentation that must carry source and review date.
  43. - `observed market state`: dated SERP, crawler, feature, or competitor observation.
  44. - `hypothesis`: testable explanation, not a finding.
  45. 5. For current platform rules, consult `data/seo-source-registry.json`. Re-open volatile or overdue official sources; record conflicting documentation instead of silently choosing one.
  46. 6. Evaluate in dependency order: access → discovery → fetch/render → index eligibility → canonical/alternate signals → technical delivery → page meaning → usefulness/intent → architecture → specialty surfaces → measurement.
  47. 7. Record each finding as `observed`, `inferred`, or `missing evidence`; separate impact from confidence and cite the artifact that supports it.
  48. 8. Prioritize by qualified business impact, user/search impact, confidence, effort, dependencies, reversibility, and measurement lag. Avoid universal SEO scores.
  49. 9. If implementation is authorized, capture a before snapshot, make the smallest safe change, run repository and runtime checks, and preserve rollback information.
  50. 10. Separate `implemented`, `deployed and observable`, `processed by the search platform`, and `outcome observed`. Preserve rerun inputs and a dated monitoring plan.
  51. ## Non-Negotiable Evidence Rules
  52. - Never invent search volume, difficulty, traffic, rankings, backlinks, conversion, competitor, crawl, or index metrics. Use `unknown` when evidence is absent.
  53. - Never guarantee crawling, indexing, ranking, rich results, AI citations, traffic, or revenue.
  54. - Do not turn title length, description length, H1 count, word count, keyword density, reading level, link count, or keyword position into universal ranking pass/fail rules.
  55. - Keep controls distinct: robots.txt governs crawler access; robots meta/X-Robots-Tag governs supported index/presentation behavior; canonical is a preference signal; sitemap and IndexNow are discovery/change notifications. None guarantees indexing.
  56. - Keep validators distinct: valid Schema.org syntax does not prove eligibility for a Google search feature; feature eligibility does not guarantee appearance.
  57. - Keep performance evidence distinct: lab tools diagnose a controlled run; field Core Web Vitals describe real-user distributions. Do not substitute one for the other.
  58. - Keep data scope visible: Search Console tables/APIs may omit anonymized queries or lower-volume rows; aggregates, filtered tables, page/query dimensions, and search types are not interchangeable.
  59. - Static HTML and rendered DOM are separate artifacts. A crawler that renders JavaScript does not prove every engine, AI bot, or user-triggered agent does so identically.
  60. - Do not attribute a traffic change to an algorithm update, migration, content change, or technical issue from timing alone.
  61. - Reject cloaking, doorway pages, scaled low-value content, expired-domain abuse, fake reviews/mentions, link spam, hidden content, and destructive bulk pruning without page-level evidence.
  62. ## Current AI-Search Boundary
  63. - For Google generative Search, foundational SEO remains the base; Google documents no special AI schema, required AI text file, ideal AI chunk size, or need to rewrite content for AI.
  64. - As of 2026-08-03, Google's newer generative-AI guide documents a dedicated Search Console Generative AI performance report while AI-feature activity also contributes to Search performance. Verify property availability and current documentation before describing reporting.
  65. - For OpenAI, distinguish `OAI-SearchBot` (Search), `GPTBot` (potential model training), and user-triggered `ChatGPT-User`; their controls are not interchangeable.
  66. - Observe Perplexity, Microsoft, and other providers independently. One prompt, citation, crawler log, or referral is a sample—not a stable ranking report.
  67. ## Gate Ladder
  68. - `Advisory`: scope, official sources, unknowns, and prioritized plan; no site-specific diagnosis.
  69. - `Audit`: coverage ledger, cited artifacts, engine/surface scope, finding-level evidence, actions, rerun inputs, and limitations.
  70. - `Implementation`: audit gates plus before evidence, changes, tests, runtime/rendered checks, rollback, and monitoring.
  71. - `Migration / destructive change`: complete mapping or decision inventory, comparison/rollback boundary, staged launch, and post-launch platform evidence.
  72. - `Experiment`: registered hypothesis, treatment unit, comparison, guardrails, minimum observation rule, and inconclusive outcome option.
  73. ## Output Contract
  74. Unless the user requests another format, provide:
  75. 1. executive summary with the top three priorities
  76. 2. outcome, scope, search surface, engine/provider, work mode, and evidence mode
  77. 3. coverage ledger and source-freshness note
  78. 4. findings: category, issue, status, evidence level/reference, impact, confidence, fix, effort, dependency, and verification
  79. 5. quick wins, strategic work, experiments, and destructive actions separated
  80. 6. keyword/page map, content brief, URL mapping, or specialty checklist only when relevant
  81. 7. implementation record and four-stage outcome status when changes were made
  82. 8. rerun inputs, monitoring window, decision rule, and rollback boundary
  83. 9. missing evidence, conflicting sources, limitations, and next measurement step
  84. For machine-readable audits, follow [Audit Contract](references/audit-contract.md) and validate with:
  85. ```bash
  86. python3 scripts/validate_audit.py path/to/audit.json
  87. ```
  88. Before relying on mutable SEO knowledge or publishing an upgrade, validate the source registry:
  89. ```bash
  90. python3 scripts/validate_knowledge.py .
  91. ```
  92. ## Qiaomu Defaults
  93. - Write concise Chinese unless the user requests another language.
  94. - Explain user and business consequences before specialist terminology.
  95. - Preserve URLs, dates, markets, devices, tools, versions, and evidence gaps for mutable claims.
  96. - Copyright (c) 向阳乔木
  97. - X: https://x.com/vista8
  98. - GitHub: https://github.com/joeseesun/
README.md · README · 135 lines

View this file on GitHub (read at the repository's current version: the scan did not record which commit it read)

  1. # qiaomu-seo
  2. > 用于 Google、Bing 与 AI Search 的专业 SEO Agent Skill:技术 SEO 审计、关键词—页面策略、自然流量诊断、站点迁移、代码修复与效果验证。
  3. [![GitHub stars](https://img.shields.io/github/stars/joeseesun/qiaomu-seo?style=flat-square)](https://github.com/joeseesun/qiaomu-seo/stargazers)
  4. [![Last commit](https://img.shields.io/github/last-commit/joeseesun/qiaomu-seo?style=flat-square)](https://github.com/joeseesun/qiaomu-seo/commits/main)
  5. [![License](https://img.shields.io/github/license/joeseesun/qiaomu-seo?style=flat-square)](LICENSE)
  6. `qiaomu-seo` 不是检查 title 字符数、H1 数量和关键词密度的打分器。它按照真实的搜索链路工作:
  7. **爬虫访问与 URL 发现 → 抓取与 JavaScript 渲染 → 可索引性(indexability)→ canonical / hreflang → 页面相关性与搜索意图 → 内链与信息架构 → 搜索表现测量。**
  8. ## 它解决哪些 SEO 问题
  9. - **页面不收录或收录异常**:检查 robots、状态码、渲染、索引指令、canonical、sitemap、内链发现与 Search Console 证据,定位阻塞发生在哪一层。
  10. - **自然搜索流量下降**:按页面、查询、国家、设备、搜索类型和时间切片,区分技术故障、需求变化、内容衰退、迁移影响与数据口径问题。
  11. - **关键词没有对应落地页**:研究搜索意图与主题关系,生成 keyword-to-page map、内容 brief、页面新建/合并/改进建议,避免关键词互相竞争。
  12. - **技术 SEO 需要落地**:审计并修复 canonical、redirect、sitemap、robots、metadata、JSON-LD、hreflang、分页、内部链接和 JavaScript 渲染问题。
  13. - **网站改版或迁移**:建立 URL inventory、重定向映射、canonical/hreflang/sitemap 一致性检查、上线监控和回滚边界。
  14. - **电商、国际化和大型站点**:检查商品变体、分类/筛选页、程序化页面、多语言/多地区信号,并按模板抽样说明实际覆盖率。
  15. - **AI Search 可见性**:分别检查 Google AI Overviews / AI Mode、ChatGPT Search、Copilot 与 Perplexity 的基础可访问性和供应商边界,不把传统 SEO、模型训练和 AI 引用混为一谈。
  16. ## 你会得到什么
  17. | 输入 | 交付物 |
  18. |---|---|
  19. | 网站 URL、代码、渲染页面、sitemap 或 crawl 文件 | 按影响、置信度、工作量和依赖排序的技术 SEO 问题清单 |
  20. | Search Console、Bing Webmaster Tools、分析数据或服务器日志 | 分段后的流量/索引诊断、候选原因、缺失证据与下一步验证 |
  21. | 产品、受众、市场和关键词数据 | 搜索意图分析、keyword-to-page map、内容 brief 与站点架构建议 |
  22. | 迁移计划或新旧 URL inventory | URL 映射、redirect/canonical/hreflang/sitemap 检查与监控方案 |
  23. | 明确授权修改的网站代码 | 最小化代码修复、变更记录、测试结果、回滚信息和复查时间点 |
  24. 每个结论都会标明它是**直接观察、合理推断还是缺少证据**。影响和置信度分开表达,不编造搜索量、关键词难度、排名、流量或竞品数据,也不承诺排名、收录、富结果或 AI 引用。
  25. ## 安装
  26. 一行安装到 Codex、Claude Code、Cursor 等兼容 Agent Skills 的工具:
  27. ```bash
  28. npx skills add joeseesun/qiaomu-seo
  29. ```
  30. 安装后可用下面的命令确认 skill 已被发现:
  31. ```bash
  32. npx skills add joeseesun/qiaomu-seo --list
  33. ```
  34. ### 前置条件
  35. - [ ] 已安装 Node.js:运行 `node --version` 检查;未安装时从 [nodejs.org](https://nodejs.org/) 获取 LTS 版本。
  36. - [ ] 可运行 npm/npx:运行 `npx --version` 检查;它会随 Node.js 一同安装。
  37. - [ ] 使用支持 Agent Skills 的客户端,并允许它读取本机 skills 目录。
  38. ## 你可以直接这样说
  39. - “抓取并渲染这个网站,找出为什么产品页没有进入 Google 索引。”
  40. - “分析这份 Search Console 导出,按页面和查询定位过去 90 天自然流量下降的原因。”
  41. - “为这个 SaaS 做关键词研究和 keyword-to-page map;没有工具数据的指标标记 unknown。”
  42. - “修复这个 Next.js 项目的 canonical、sitemap、robots 和 Product JSON-LD,并验证渲染结果。”
  43. - “为网站迁移生成完整 URL redirect map,检查 canonical、hreflang 与 sitemap 是否一致。”
  44. - “审计多语言电商站的商品变体、faceted navigation、分页和 hreflang。”
  45. - “分别检查 Google AI Overviews、ChatGPT Search 和 Perplexity 的抓取与可见性基础,不承诺引用。”
  46. - “为这批内容页设计 SEO 实验,定义处理组、对照组、观察窗口和停止规则。”
  47. ## 专业边界
  48. - **审计 / 诊断 / 比较**默认只读,不修改代码、索引控制、站长平台或线上数据。
  49. - **修复 / 实现**只在明确授权后执行,并保留修改前证据、测试结果、回滚信息和监控窗口。
  50. - robots.txt 控制抓取访问,robots meta / X-Robots-Tag 控制受支持的索引与展示行为,canonical 是偏好信号,sitemap 与 IndexNow 是发现/变更通知;它们都不保证收录。
  51. - Schema.org 语法有效、Google 搜索功能资格和实际出现富结果是三个不同层级。
  52. - Lighthouse 等实验室数据用于诊断;Core Web Vitals 现场数据描述真实用户分布,二者不互相替代。
  53. - 不覆盖付费搜索投放、App Store Optimization、垃圾外链、隐藏内容、关键词堆砌或排名保证。
  54. ## 1.2 专业能力升级
  55. - 登记并定期复查 33 个 Google、Bing、IndexNow、Schema.org、web.dev、OpenAI、Microsoft 和 Perplexity 官方来源。
  56. - 对平台规则记录来源、复查日期和功能生命周期,自动拦截已知陈旧断言。
  57. - 分离 Schema.org 语法、搜索功能资格和实际展示;分离 IndexNow 接收、抓取与索引结果。
  58. - 增加国际 SEO、电商、图片/视频搜索、程序化 SEO、内容治理、性能测量和 SEO 实验模块。
  59. - 区分 OAI-SearchBot、GPTBot、ChatGPT-User 与其他 AI crawler 的用途和控制边界。
  60. 设计依据、竞品 skill 取舍与验证记录见 [`reports/creation-handoff.md`](reports/creation-handoff.md)。
  61. ## 验证
  62. ```bash
  63. python3 scripts/validate_skill.py .
  64. python3 scripts/validate_knowledge.py . --strict-stale
  65. python3 scripts/validate_audit.py path/to/audit.json
  66. python3 ~/.agents/skills/qiaomu-meta-skill/scripts/export_skill_ir.py . --output reports/skill-ir.json
  67. python3 ~/.agents/skills/qiaomu-meta-skill/scripts/trigger_eval.py . --cases evals/trigger_cases.json --output reports/trigger-eval.json
  68. ```
  69. ## Troubleshooting
  70. | 问题 | 处理 |
  71. |---|---|
  72. | 没有 Search Console 或关键词工具 | 做定性研究,把搜索量、KD、覆盖率标记为 `unknown` |
  73. | 静态抓取看不到 JSON-LD | 检查渲染后的 DOM,或使用 Rich Results Test |
  74. | 想知道是否被索引 | 优先使用 Search Console URL Inspection;不要把 `site:` 查询当完整索引报告 |
  75. | 只想审计,不想改站 | 使用“审计/诊断/比较”措辞;skill 会保持只读 |
  76. | 官方规则可能变化 | 查看 `data/seo-source-registry.json` 并运行 `validate_knowledge.py` |
  77. | Schema 验证通过但无富结果 | 分别检查语法、当前功能资格、页面政策和实际展示 |
  78. | Lighthouse 满分但 CWV 未通过 | Lighthouse 是实验室诊断;检查 CrUX、Search Console 或真实用户现场数据 |
  79. ## 版权
  80. Copyright (c) 向阳乔木
  81. X: https://x.com/vista8
  82. GitHub: https://github.com/joeseesun/
  83. <!-- qiaomu-profile:start -->
  84. ## 关于向阳乔木
  85. 向阳乔木(乔向阳 / Joe)是一位实践型 AI 产品与内容创作者,长期把前沿 AI 变化转译成可复用的工作流、产品判断、AI 编程实践、AI 搜索实践和 GEO/AI 营销方法。
  86. - 个人网站: https://qiaomu.ai
  87. - 博客: https://blog.qiaomu.ai
  88. - X: https://x.com/vista8
  89. - GitHub: https://github.com/joeseesun/
  90. - 微信公众号: 向阳乔木推荐看
  91. ### 支持与关注
  92. | 打赏支持 | 微信公众号 |
  93. |---|---|
  94. | <img src="assets/qiaomu-profile/qiaomu_reward_qr.png" alt="向阳乔木打赏二维码" width="180" /> | <img src="assets/qiaomu-profile/qiaomu_wechat_public_account_qr.jpg" alt="向阳乔木推荐看公众号二维码" width="180" /> |
  95. | 感谢支持乔木持续分享 AI 实践 | 扫码关注「向阳乔木推荐看」 |
  96. <!-- qiaomu-profile:end -->
  97. ## License
  98. MIT

Ratings and comments

Sign in to rate this skill.

Sign in

Free to use. If it’s helped, pass it on.