gingiris-skills

Iris / Gingiris 的出海增长工具箱 —— 把 6 个 gingiris-* playbook repo + 每日运营脚本 + 60 篇博客实战经验，打包成 Claude Code skills。

作者：Iris · Gingiris 博客

所有内容开放，可以整套装，也可以只拿一部分。Skill、知识包、原子库、单个脚本都能单独用。

工具箱（v0.4.0 — 12 个 skill）

| /gr-backlinks | 系统化外链建设 — Wikipedia / PR-HARO / G2 / Reddit-Quora 5 通道，0→1 站点 GEO + SEO 共升 |

Roadmap（0.3+）

工作流

gr-competitor（看对手在做啥）
    ↓
gr-ph-launch / gr-oss-marketing / gr-b2b（选打法）
    ↓
gr-blog-post（产内容）
    ↓
gr-seo-patrol（上线后监控）
    ↓ cannibalization           ↓ 雪崩
gr-seo-patrol canonical-fix   gr-seo-patrol rescue
    ↓
gr-user-interview（用户反馈）

Skill 之间会自动推荐下一步。比如：

• gr-ph-launch 发布 24h 后 → 推荐 gr-seo-patrol 加监控
• gr-seo-patrol 发现 cannibalization → 自动跳到 canonical 修复流程
• gr-blog-post 发布后 → 自动加入 gr-seo-patrol 监控名单

安装

推荐：Claude Code 插件市场

claude plugin marketplace add Gingiris/gingiris-skills
claude plugin install gr@gingiris-skills

装完在 Claude Code 里输入 /gr 即可。

单独装某个 skill：

claude plugin install gr-seo-patrol@gingiris-skills

知识库

所有方法论文档与原子知识点都开放，即使不装 skill 也能用。

目录结构

知识库/
├── 原子库/
│   ├── atoms.jsonl                    # 结构化知识原子（可 RAG）
│   └── README.md
└── Skill知识包/
    ├── iris_writing_style.md          # 文风指南（5 要素）
    └── seo_geo_playbook_2026.md       # SEO 飞轮 + GEO 三件套

怎么在你自己的项目里用

场景 1：给你的 AI 加 SEO 诊断能力 把 知识库/Skill知识包/seo_geo_playbook_2026.md 粘到 system prompt。

场景 2：做 RAG atoms.jsonl 导向量库，每条带 topics 标签。

场景 3：只要脚本 skills/gr-seo-patrol/scripts/*.py 可独立跑，看 docs/api-keys-template.md 配 env。

API 依赖

见 docs/api-keys-template.md。

核心：DATAFORSEO_B64 + GITHUB_TOKEN，其他 skill 按需加。

许可证

MIT。

• 个人使用、学习、研究：随便
• 商业用途：随便
• 衍生作品：建议（不强制）注明来源

Monthly Full-Site Audit Workflow

Battle-tested 2026-05-07 on a 58-page blog. Caught 43 SERP-truncating titles + 36 schema warns + 27 stop-word slugs in a single 30-min run. One layout-level commit fixed 20 of 43 titles. Use for any Jekyll / Hugo / Next.js blog with 30+ posts.

A repeatable 6-stage workflow you can run on any site. Powered by 4 scripts (see attribution below).

Stage 1 — Discovery (5 min)

Pull all blog URLs from your sitemap:

import urllib.request, re
sm = urllib.request.urlopen("https://your-site.com/sitemap.xml").read().decode()
urls = [u for u in re.findall(r"<loc>([^<]+)</loc>", sm) if "/blog/" in u]

Stage 2 — Parallel Audit (20 min for 60 pages)

Run two audit scripts per URL in 4-thread parallel:

pip install requests
python3 skills/gr-seo-patrol/scripts/check-page.py URL --timeout 20
python3 skills/gr-seo-patrol/scripts/check-schema.py URL --timeout 20

Each script outputs a structured JSON envelope (status: pass|warn|fail|info per check).

Stage 3 — Aggregate Findings

Bucket issues by type:

• Title length > 70 chars (SERP truncation risk)
• H1 length > 70 chars (mobile readability)
• Meta description outside 80-170 chars
• Schema warns by @type (BlogPosting / Article / Organization)
• Canonical mismatches, slug stop words, missing alt text

Save aggregated counts + per-URL lists to findings.json.

Stage 4 — Layered Fix Strategy (HIGH ROI ORDER)

Stage 5 — Verify

After Jekyll/Hugo rebuild (~60-90s), re-run check-schema.py on a sample page. All schema types should show status: pass: Article · BlogPosting · Organization · FAQPage

Stage 6 — Archive + Trend Track

Commit findings.json to data/audit-{YYYY-MM-DD}.json for month-over-month trend analysis. Add 2-5 atoms to 知识库/原子库/atoms.jsonl documenting any new lessons.

Schedule it

Add a monthly cron to auto-run before your Phase 2 / quarterly checkpoint:

# In Claude Code's scheduled-tasks
cronExpression: "0 10 1 * *"   # 10am on day 1 of each month
prompt: "Run Monthly Full-Site Audit per gr-seo-patrol/SKILL.md workflow..."

What you'll typically find on your first run

Real numbers from gingiris.github.io/growth-tools 2026-05-07 run:

Total time: ~30 min audit + 90 min fixes = 2 hours for site-wide SEO health refresh.

HARD RULE (anti-hallucination guardrail)

⛔ Output ONLY the checks defined in the script's JSON envelope.

• Do NOT add "bonus" checks not in the script output
• Do NOT contradict the script's status field without observable evidence
• Do NOT invent metrics like "EEAT score 89" — third-party scoring is unofficial per Google 2026 guidance
• If llm_review_required: true, make explicit judgment + document reasoning + update status

The script envelope is the single source of truth. Treat as strict whitelist.

Script attribution

The 4 audit scripts (check-page.py, check-schema.py, check-site.py, check-social.py) in skills/gr-seo-patrol/scripts/ are adapted from JeffLi1993/seo-audit-skill (MIT). Original repo focused on single-page client-presentable HTML reports; we adapted them for orchestrated batch audit + Jekyll/GitHub Pages site analysis. Original license terms preserved in each file header.

关联 repo

• Gingiris/growth-tools — 主 blog，所有 skill 的实战场
• Gingiris/gingiris-launch — PH launch playbook 内容源
• Gingiris/awesome-agent-skills — skills 生态索引
• dontbesilent2025/dbskill — 框架灵感来源
• JeffLi1993/seo-audit-skill — 单页 SEO 审计脚本来源（4 scripts adapted）
• AgriciDaniel/claude-seo — 25-skill SEO 全套（含外链分析，可作 Tier 3 扩展）
• zubair-trabzada/geo-seo-claude — GEO 0-100 评分体系（可作 gr-geo-cite 增强）

贡献

• 发 Issue 描述你的运营场景
• PR 加新 skill / 新原子 / 新方法论
• 跟 @WeiYipei 聊