agentleFS
RomeoSEORomeoSEO/seo-skills/skills/seo-opportunity-research/SKILL.md

seo-opportunity-research

Skill7 starsChanged 14 days agoMITView on GitHub

Find and score SEO opportunities from your own rankings, keyword data, SERPs, competitors and site inventory, then decide create, refresh, consolidate, link or skip. Use for "what should I write next", "keyword research for my site", "find page 2 keywords", "which keywords are worth targeting".

The file

Show its rules and text

We split this skill into the rules it gives an agent

16 rules. About 0 are common, 14 are distinctive, and 2 are very specific to this project.

  • ≈0

    Common · estimate

    Many files like this have a rule of this kind.

  • ≈14

    Distinctive · estimate

    This author's own choices. Worth a look.

  • 2

    Very specific

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

  • Common
  • Distinctive
  • Very specific
  1. Distinctive:

    Search performance data when available: query, page, clicks, impressions, CTR, average position, country, device, date.

    line 41
  2. Distinctive:

    Do not create content that competes with a stronger existing page unless the recommendation is consolidation or refresh.

    line 66
  3. Distinctive:

    Do not report the same unchanged opportunity repeatedly.

    line 67
  4. Distinctive:

    Keep each source's quirks out of the analysis.

    line 80
  5. Distinctive:

    Do not pursue: poor business fit, wrong SERP intent, too broad, too competitive, or already sufficiently covered.

    line 95
  6. Distinctive:

    Start from business themes, product use cases, customer roles, platforms, competitors, and pain points.

    line 121
  7. Distinctive:

    Prefer long-tail 100-1000 volume for realistic small-site wins unless the site has authority for larger terms.

    line 122
  8. Distinctive:

    Check whether the site already ranks.

    line 123
  9. Distinctive:

    If SERP intent is polluted or mismatched, downgrade or mark do not pursue.

    line 132
  10. Distinctive:

    Search demand: 0-5. Use volume, impressions, trend, and seasonality.

    line 137
  11. Distinctive:

    Create only when no existing page satisfies intent.

    line 145
  12. Distinctive:

    Update snippet only when the current title/meta is clearly weak and the page is not a sensitive ranked page.

    line 149
  13. Distinctive:

    Check site inventory, active tasks, recent reports, and completed work.

    line 155
  14. Distinctive:

    If already tracked, return the existing item and do not create a duplicate recommendation.

    line 157
  15. Very specific:

    Run seo-content-brief on the top create, refresh, landing or comparison recommendation, passing its evidence, score, intent and SERP notes.

    line 229
  16. Very specific:

    If site coverage was thin or missing, run site-content-inventory first, then re-run this skill.

    line 231

≈ “Common” is an estimate: at least 1 in 5 of the 364,980 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 opportunity research

Decide what SEO work should happen next: create new content, refresh a page, consolidate overlapping pages, improve internal links, adjust snippets, build a landing page, or ignore a tempting but weak keyword. This is not a generic keyword research skill. It is an evidence-weighted opportunity engine. Keyword volume alone is never enough. It hands back a ranked, de-duplicated list of opportunities, each with a type, evidence, a score, a recommended action and the next skill to run.

Romeo's rule: The keyword with the biggest number is usually the one you can't have; the one you already rank 14th for is the one you can.

Use this when

  • You need to decide what SEO work to do next on a site, ranked by evidence rather than volume.
  • You have Search Console data or a keyword export and want the page-2, low-CTR and long-tail wins pulled out of it.
  • You have a pile of keyword ideas and need to know which deserve a new page, which belong on an existing page, and which to ignore.
  • A commercial, platform or competitor query might deserve a landing or comparison page instead of a blog post.
  • You are researching a non-English market and need native keyword behavior, not translations.

Don't use this for

  • Explaining why traffic rose or dropped → seo-performance-audit.
  • Grouping a large keyword set into hub-and-spoke clusters → topic-cluster-map.
  • Turning a chosen opportunity into a writer-ready plan → seo-content-brief.
  • Deciding merge vs differentiate for pages already competing → keyword-cannibalization-check.

What you need

Required

  • Site domain and canonical URL rules.
  • Target market: countries, languages, regions, excluded locales.
  • Business profile: product category, ideal customers, use cases, pricing model, conversion path.
  • Product facts and prohibited claims.
  • Existing site inventory, crawl, sitemap, CMS export, or uploaded URL list.
  • Search performance data when available: query, page, clicks, impressions, CTR, average position, country, device, date.
  • Keyword provider data when available: volume, difficulty/competition, trend, SERP URLs, related terms.
  • Competitor domains and comparison targets.
  • Topics to prioritize, avoid, or suppress.
  • Known active tasks and recent SEO work, so you do not repeat yesterday's work.

Helpful

  • Analytics traffic/conversion by URL.
  • Revenue or lead-quality events.
  • Prior SEO reports.
  • Editorial calendar.
  • Support tickets, sales-call notes, product usage, internal search, or AI-bot/server-log demand.
  • Localization glossary and country-specific keyword rules.

No tools?

You can still run this at a basic level. A Search Console Performance export (Queries and Pages, last 3 months compared with the previous 3, as CSV) covers the existing-ranking step on its own: page-2 positions, high-impression low-click rows and long-tail wins. Build site coverage from /sitemap.xml opened in a browser, or a URL list the user pastes. For demand, use Google Keyword Planner (free with a Google Ads account; it shows ranges), Google Trends for direction and seasonality, and autocomplete and "People also ask" for phrasing. Inspect the SERP yourself in a private window set to the target country, or ask the user for screenshots. Any volume or difficulty you could not measure is "not checked"; never estimate it.

Ground rules

  • First-party data wins over generic keyword suggestions.
  • Existing ranking pages are often more valuable than new articles.
  • Page-2 rankings with business fit are the fastest wins.
  • Long-tail opportunities beat broad vanity terms for small and mid-sized sites.
  • Do not create content that competes with a stronger existing page unless the recommendation is consolidation or refresh.
  • Do not report the same unchanged opportunity repeatedly. Preserve suppression windows and already-tracked issue checks.
  • For localized SEO, research native keyword behavior. Never translate slugs or keywords literally.

Data sources

Accept normalized data from whatever sources are available. Examples:

  • Google Search Console for existing query/page performance.
  • DataForSEO, Ahrefs, Semrush, Moz, Keyword Planner, or CSV for keyword/competitor data.
  • Browser automation, SearchAPI, SerpAPI, or cached SERP snapshots for live SERP inspection.
  • CMS, crawler, or sitemap export for inventory.
  • Analytics export for traffic/conversion weighting.

Keep each source's quirks out of the analysis. Normalize everything into compact records first, then reason over those records.

Opportunity types

Classify every finding as exactly one primary type:

  • Page-2 push: ranking positions 11-30 with relevant demand.
  • CTR lift: high impressions, weak CTR, page already matches intent.
  • Content refresh: page matches intent but is stale, thin, badly structured, or missing answer-first sections.
  • Internal-link lift: relevant page ranks or converts but lacks links from supporting pages.
  • New content: no existing page matches the intent.
  • Landing page: commercial/platform/tool intent deserves a conversion page rather than a blog post.
  • Comparison page: competitor or alternative intent deserves a decision page.
  • Consolidation: duplicate intent or cannibalization exists.
  • Localization opportunity: local-language demand differs from English targeting.
  • Do not pursue: poor business fit, wrong SERP intent, too broad, too competitive, or already sufficiently covered.

Steps

  1. Build the business/search context.
    • Summarize the product, buyer, main use cases, and conversion goals.
    • Identify branded terms, non-branded priority themes, comparison terms, platform terms, pain-point terms, and forbidden topics.
    • Load locales and exclusions. For example, a site owner may want English-only reporting, or may want German research to use native German terms.
    • Identify conversion URLs and pages that should receive internal links.
  2. Build or load site coverage.
    • Normalize URLs and remove tracking parameters.
    • Group pages by section: blog, landing page, comparison, docs, help, product, pricing, templates, glossary.
    • Attach known target keywords and current page purpose.
    • Mark thin pages, stale pages, uncategorized pages, and pages with no conversion path.
  3. Mine existing ranking opportunities.
    • Pull recent search performance with a comparison period when available.
    • Exclude configured noise: irrelevant locales, test URLs, suppressed clusters, branded terms when non-branded research is requested.
    • Find high-impression low-click records. Default signal: >200 impressions, <5 clicks, CTR <1%, adjusted for site size.
    • Find page-2 opportunities. Default signal: average position 11-30, >100 monthly volume when volume exists, or meaningful impressions when volume is absent.
    • Find long-tail hidden wins. Default signal: 3-4+ words, relevant, already ranking or generating impressions.
    • Find scaling pages. Default signal: CTR >3% with low impressions, or high engagement/traffic with low search visibility.
    • Detect cannibalization by grouping queries/pages by intent, not just exact keyword.
  4. Mine new keyword opportunities.
    • Start from business themes, product use cases, customer roles, platforms, competitors, and pain points.
    • Prefer long-tail 100-1000 volume for realistic small-site wins unless the site has authority for larger terms.
    • Check whether the site already ranks. If rank is 1-10, classify as defend/scale, not new content. If 11-30, classify as page-2 push. If >30 or absent, consider new/refresh.
    • Filter out one-word/generic terms, wrong-language terms, polluted intent, and topics that require claims the site cannot support.
  5. Inspect the SERP before recommending content.
    • Review top 3-5 ranking pages or SERP records.
    • Capture page type: blog, listicle, landing page, product page, support doc, template, tool, forum, marketplace, video.
    • Note answer pattern: quick answer, steps, table, feature grid, calculator, FAQ, comparison, local pack, forum-heavy, UGC-heavy.
    • Identify what ranking pages do well: title angle, headings, depth, examples, tools covered, media, schema.
    • Identify gaps the site can credibly fill.
    • If SERP intent is polluted or mismatched, downgrade or mark do not pursue.
  6. Score each candidate.
    • Business fit: 0-5. Does this attract people who can become customers?
    • Evidence strength: 0-5. First-party ranking/impression/conversion evidence beats external keyword guesses.
    • Search demand: 0-5. Use volume, impressions, trend, and seasonality.
    • Rankability: 0-5. Consider current rank, domain strength, SERP type, competitor quality, page type match.
    • Conversion path: 0-5. Can the page naturally link to or become a conversion page?
    • Freshness/urgency: 0-5. Rising demand, content decay, recent drop, or urgent commercial opportunity.
    • Effort: 0-5 reversed. Refresh/internal link is easier than a full new page; technical/design dependencies reduce score.
    • Risk: 0-5 reversed. Existing ranked title/meta changes, legal/medical claims, or SERP mismatch reduce score.
  7. Decide the action.
    • Create only when no existing page satisfies intent.
    • Refresh when a page is already close but weak.
    • Consolidate when multiple pages split the same intent.
    • Internally link when the content is good but unsupported.
    • Update snippet only when the current title/meta is clearly weak and the page is not a sensitive ranked page. Protect pages already ranking well.
    • Build landing page when the query is commercial or platform/tool-specific and conversion intent is clear.
    • Build comparison page when the query implies alternatives, vs, competitor, pricing, or selection criteria.
    • Ignore when the keyword is tempting but strategically weak.
  8. De-duplicate and suppress repeats.
    • Check site inventory, active tasks, recent reports, and completed work.
    • Compare by URL, keyword cluster, and intent, not exact title.
    • If already tracked, return the existing item and do not create a duplicate recommendation.
    • Suppress repeated clusters unless there is a material new event: large movement, new keyword outside the cluster, technical issue, or explicit user request.

Quality bar

A good recommendation should answer:

  • Why this opportunity matters now.
  • Whether to create or improve an existing page.
  • Which intent the page must satisfy.
  • What proof supports the recommendation.
  • What not to do.

What to hand back

Fill in this template. Rank recommendations by score, highest first.

# SEO opportunities: <site> (<market>, <data date range>)

**What I found:** <one plain-English sentence>
**What matters most:** <one sentence>
**What to do first:** <one sentence>

## Executive summary
- <3-6 bullets with the highest-signal findings>

## Ranked recommendations

### 1. <keyword cluster>
- Opportunity type: <exactly one primary type>
- Keyword cluster: <primary query + close variants>
- Current/proposed URL: <URL>
- Target country/language: <...>
- Evidence: <search / keyword / SERP / inventory facts, each marked seen, tool or inferred>
- Score: <total>/40 (business fit x, evidence x, demand x, rankability x, conversion path x, freshness x, effort x, risk x)
- Confidence: high / medium / low
- Priority: `Fix now` / `Fix next` / `Worth doing` / `Leave it`
- Recommended action: <create / refresh / consolidate / internal link / snippet update + risk note / landing page / comparison page / ignore>
- Why this matters for the business: <...>
- What not to do: <...>
- Next skill to call: <`seo-performance-audit` / `seo-content-brief` / `seo-blog-writer` / `seo-content-qa` / `site-content-inventory` / `cms-publishing-adapter` / `comparison-page-writer` / `internal-linking-plan` / `title-meta-ctr` / `keyword-cannibalization-check`>

### 2. ...

## Already covered or suppressed
- <cluster>: <existing URL or tracked item>, <why it is not a new recommendation>

## Do not pursue
- <keyword or cluster>: <reason>

## Data gaps that would improve confidence
- <missing source and what it would change>

## What I could not check
- <e.g. search volume not checked: no keyword data provided>

Stop and flag if

Fail the research, and say so instead of handing it back as done, if:

  • It proposes topics from volume alone.
  • It ignores existing pages and creates duplicate content.
  • It misses obvious first-party page-2 or low-CTR opportunities.
  • It reports excluded languages/markets.
  • It repeats unchanged findings without checking prior work.
  • It recommends title/meta changes for already-ranking pages without a risk note.
  • It treats translated keywords as valid without local keyword research.

Next step

  • Run seo-content-brief on the top create, refresh, landing or comparison recommendation, passing its evidence, score, intent and SERP notes.
  • Route other types to their specialist: internal-link lifts to internal-linking-plan, CTR lifts to title-meta-ctr, consolidation to keyword-cannibalization-check.
  • If site coverage was thin or missing, run site-content-inventory first, then re-run this skill.

Part of the free RomeoSEO skill library: https://romeoseo.com/skills

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.