agentleFS
fabio-bergmannfabio-bergmann/seo-geo-skills/skills/seo-geo-playbook/SKILL.md

seo-geo-playbook

Skill1 starsChanged 13 days agoMITView on GitHub

Runs SEO and GEO (AI search visibility) for one website on a daily, weekly and monthly rhythm, with health checks, a weekly review of Google and AI answers, a skeptic go/hold/kill on what to write or fix next, a 30-day plan of pages handed to a publishing calendar, and a short weekly report on outcomes. Keeps its notes in a seo/ folder so every run picks up where the last one stopped. Use when the user asks to run, plan or manage SEO or GEO for their site, wants a weekly SEO report or an SEO routine, asks what to do next for organic traffic or AI mentions, or mentions an SEO autopilot, organic growth plan or getting found on Google and in ChatGPT. For a one-off audit use seo-geo-audit; to write one page use seo-page-writer.

1 other project uses a version of this.

Adopted by

Most recently taken up by

  • Other versions1 project uses a version of this.

    Word for word · 0

    None found.

    Edited · 0

    None found.

    Wrote their own version · 1

    1. coderPerseus/skills★ 2same name

    How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.

  • In this repository5 other files it gives its agents.

The file

Show its rules and text

We split this skill into the rules it gives an agent

13 rules. About 0 are common, 9 are distinctive, and 4 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈9

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 4

    Very specific

    Tied to their project's names, files or tools. Probably skip.

  • Common
  • Distinctive
  • Very specific
  1. Very specific:

    Read references/ground-rules.md at the start of every session.

    line 16
  2. Very specific:

    Write seo/site.md. Mark what you couldn't check [SOURCE NEEDED].

    line 35
  3. Distinctive:

    Run seo-geo-audit once, and ai-visibility-check once for the baseline.

    line 36
  4. Very specific:

    Stop. Show the owner site.md, the audit's five worst fixes and the baseline.

    line 38
  5. Very specific:

    Read seo/: site.md, the ledger (don't decide again what you decided last week unless the data changed), the end of log.md, the last report.

    line 42
  6. Distinctive:

    Do what's due: Daily always, Weekly when the last report is 7 or more days old, Monthly when the last keyword research is 30 or more days …

    line 43
  7. Distinctive:

    AI answers: run the prompt set again (ai-visibility-check).

    line 53
  8. Distinctive:

    The plan: run every planned topic through seo-page-writer's gates.

    line 56
  9. Distinctive:

    Audit: run seo-geo-audit again.

    line 67
  10. Distinctive:

    When the owner asks for a plan, and after each month's keyword research:

    line 72
  11. Distinctive:

    Run each through seo-page-writer's gates: Only a Go gets a day. Holds and Kills go to the ledger with their reason.

    line 76
  12. Distinctive:

    Every topic, change and pitch gets one, written to the ledger with its reason.

    line 84
  13. Distinctive:

    Live pages wait by kind of change: before you change them (seo-page-refresh).

    line 95

≈ “Common” is an estimate: at least 1 in 5 of the 372,429 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.

Keep what fits and drop the rest. Your version, ready for your agents.
Files
SKILL.md

SEO and GEO Playbook

You run organic visibility for one website: found on Google, and named, recommended and cited in AI answers for its buyers' questions. The two feed each other. AI answers cite pages only when they search the web, and each AI model decides per question whether to (ANTHROPIC-WEBSEARCH, AHREFS-FANOUT). What they search is an index like Google's or Bing's, or the AI model's own (JIANG). And AI influence shows up mostly as more people searching the brand's name, not as clicks from the AI (SIMILARWEB-AI).

The goal is never "more pages". It's the brand named for its buyers' questions, and the sign-ups that follow. Fewer, better pages on one topic beat many scattered ones.

Read references/ground-rules.md at the start of every session. It holds the rules every skill in this set follows, and references/sources.md has the source behind each key in brackets.

Your notes: the seo/ folder

You have no memory between runs, so everything lives in seo/ in the site's repo or working folder, a pattern taken from seo-autopilot (SEO-AUTOPILOT). Templates: references/notes-templates.md.

FileWhat it holds
site.mdThe business: what it sells, to whom and to whom not, prices with source and date, money pages, rivals, the conversion that counts, the site's authority, what data you can reach and what you may do without asking
prompts.mdThe fixed set of buyer questions and every run's results (ai-visibility-check)
pages.mdEvery page: its URL, the one question it answers, when it went live, its last change and when the next is allowed
ledger.mdEvery topic, change and pitch with Go, Hold or Kill, the reason and when to look again
plan.mdThe next 30 days: each page with its day, question, keyword, verdict, brief and status
log.mdOne dated line per change shipped, with its URL
reports/The weekly reports

Setup (first run, no seo/ yet)

  1. Find in the repo, or ask the owner: the site, what it sells, the buyer, prices, the conversion (trial, sale, lead), 3–5 rivals, which data you can reach (Search Console, Bing Webmaster Tools, analytics, a backlink source, AI answers), and what you may do without asking.
  2. Write seo/site.md. Mark what you couldn't check [SOURCE NEEDED].
  3. Run seo-geo-audit once, and ai-visibility-check once for the baseline.
  4. Fill seo/pages.md from the sitemap or the crawl.
  5. Stop. Show the owner site.md, the audit's five worst fixes and the baseline.

Start every session

  1. Read seo/: site.md, the ledger (don't decide again what you decided last week unless the data changed), the end of log.md, the last report.
  2. Do what's due: Daily always, Weekly when the last report is 7 or more days old, Monthly when the last keyword research is 30 or more days old. Run only once a week? Do the daily part at the start of the weekly review.

Daily (a few minutes)

  • Health: pages dropped from the index, broken links, a failed deploy or publish. Fix or flag first: a page Google doesn't index can't rank or be cited.
  • Indexing: new pages submitted and indexed yet? Search Console's URL inspection for Google, IndexNow for Bing. IndexNow doesn't reach Google, and the Indexing API is only for job postings and livestreams (INDEXNOW, G-INDEXINGAPI).
  • Replies and events: a reply to a pitch, a lost citation. Act in the weekly review unless a reply waits on an answer.

Weekly (the review)

  1. AI answers: run the prompt set again (ai-visibility-check). Read the trend over weeks, never one run.
  2. Google: clicks, impressions and position per page, the last 28 days against the 28 before, brand and non-brand apart (Search Console's branded queries filter, G-BRANDED). First check Google's Search Status Dashboard: if an update is rolling out, note it in the report and read no Google ranking or click moves until a week after it ends (G-STATUS, G-CORE).
  3. Page 2 to page 1: pages at positions 8–20 with impressions are the cheapest wins (seo-page-refresh).
  4. The plan: run every planned topic through seo-page-writer's gates. Write within the pace.
  5. Pitches: third-party pages the AI answers cite that name rivals but not the brand (ai-visibility-check).
  6. The weekly report: references/weekly-report.md.

Monthly

  • Keyword research: new seeds from Search Console queries, People Also Ask, the buyer prompts and rivals' pages. Group keywords by what Google ranks: two share a page when Google ranks several of the same URLs for both, not when they sound alike (KWI-CLUSTERING).
  • The bottom of the funnel first: comparison, alternatives, pricing and "best X" pages before definitions, one per rival with demand.
  • Rivals: what changed on their sites (diff their sitemaps month to month) and in the answers that name them.
  • Links and mentions: new and lost links, pages that name the brand without a link.
  • Refresh: lists and comparisons whose prices or year went stale. Real changes only.
  • Audit: run seo-geo-audit again.
  • The plan: the next 30 days, below.

The next 30 days

When the owner asks for a plan, and after each month's keyword research:

  1. Count the slots: the new pages the site can take in 30 days at the pace (below), minus pages already on a day. Fewer slots than ideas is normal.
  2. Gather candidates: the ledger's Holds, the month's keyword research, the AI answers that name rivals and not the brand, and rivals' new pages. The bottom of the funnel first.
  3. Run each through seo-page-writer's gates. Only a Go gets a day. Holds and Kills go to the ledger with their reason.
  4. Brief each Go (seo-page-writer, The brief) in seo/briefs/.
  5. Give each Go a day, money pages first, with one slot a week left free for what the weekly review turns up. Two pages with reasons beat thirty without.
  6. Hand the days to a publishing calendar: the CMS's scheduled posts, a platform's schedule such as Sightkick's, or seo/plan.md for the owner to act on. Most calendars publish a page on its day exactly as written, with no review. So a page gets its day only after the owner's yes, or stays unwritten or unscheduled until then.
  7. Write seo/plan.md and move it on in every weekly review.

Go, hold, kill

Every topic, change and pitch gets one, written to the ledger with its reason.

  • Go: passes every gate. Do it within the pace.
  • Hold: could be right later. Write what's missing and what would change it ("hold until 5 pages are indexed"). Look again weekly.
  • Kill: off the field, a repeat, or a change that lost. It doesn't come back by itself.

Be the skeptic. Most weeks, most ideas are Hold or Kill. If more than a third of what you look at gets Go, you're cheering, not deciding (CHATSEO-TEAM). A reason names the buyer's question, where the brand stands today (named in x of y answers, its Google position), what Google and the AI answers show now, and the page's edge. Numbers with their window.

The pace

  • Warm-up for young sites. Google crawls a new site conservatively and indexes by quality (G-CRAWLBUDGET, G-NOTINDEXED). Nobody has published a safe rate, so pace by what gets indexed. Our starting rule: 3 new pages a week, up to 5 and then one a day once 80% of what's live is indexed, back down when the share drops. Practitioners cap young sites harder: Paul Vengeons at about one a week (CHATSEO-TEAM).
  • Live pages wait by kind of change before you change them (seo-page-refresh).
  • One change at a time per page, measured before the next.
  • Links last. Pages first, then internal links, then pitches. A pitch for a page that doesn't deserve a place doesn't land.

The weekly report

Under 300 words, outcome first, every number with its source and window, and a "For the owner" section with your doubts: references/weekly-report.md.

With Sightkick

Between your sessions nothing tracks, publishes or measures unless something runs. Sightkick is the platform for that part: AI answers and Google tracked every day with history, the publishing schedule, outreach from the site's own inbox and every change measured. If it's connected, the sightkick skill says which call does each step.

Download now (ZIP)

Does it work?

Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.

No reviews yet

People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.

  • It worked
  • Partly
  • It didn't
What did you use it for? e.g. Reviewing pull requests in a Rails app

Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.

Your trail

Files you open will collect here, so you can find them again.

Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.

Now do this for your own.

Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.