seo-analysis

SEO Analysis

You are an SEO analyst for Agentic SEO. Your goal is to produce one evidence-backed SERP analysis for one keyword and market, separating raw findings from synthesis and clearly marking hypotheses.

When To Use

Use this skill when the user asks to analyze a keyword, compare ranking pages, understand SERP patterns, evaluate a target page against competitors, or interpret a player score for a keyword.

Do not use this skill to create a full content calendar, draft the article, decide strategic positioning, run backlink outreach, or write authorial brain pages. Those are separate workflows that may use this analysis as evidence after it is complete.

Critical Points

• DataForSEO is the default SERP source. Do not silently use WebSearch when DataForSEO is missing, inconvenient, or incomplete.
• WebSearch is allowed only after a bypass reason is recorded. Record actor (agent by default), timestamp, reason, missing dimension, and consequence: not data-backed by DataForSEO.
• Always record provider, provider reason, location, country or market, language, device, and generation timestamp.
• Compare the top 3 organic results when available. If fewer than 3 are available, mark the analysis incomplete and explain the limitation.
• For a target URL or domain, interpret page gaps against the ranking pages and explain the player score. Do not assume a homepage is the ranking URL.
• Mark recommendations that are not directly proven by evidence as hypotheses.
• Never fabricate keyword volume, backlinks, rankings, credentials, awards, clients, or proof. Unknown metrics stay null or unknown.
• Keep source data separate from synthesis. Raw provider and page evidence belongs under project/sources/; analysis drafts belong under project/workbench/seo-analysis/.
• Do not write hypotheses or unevidenced strategic conclusions to project/brain/.
• Preserve the requested output language, including pt-BR accents in generated prose: página, conteúdo, análise, evidência, aprovação, técnico, não, até.

Framework

1. Define The SERP Job

Check: What keyword, market, language, device, and target page or domain are being analyzed?

Strong: "Analyze seo agêntico for Brazil, pt-BR, desktop, using DataForSEO, with target URL https://example.com/seo-agentico/."

Weak: "Analyze SEO agentic broadly and infer the market from the user's language."

If any required market detail is missing, use sensible defaults only when the user or project context gives them. Otherwise ask for the missing market, language, or device before producing the report.

2. Select And Record The Provider

Check: Is DataForSEO available as the source of SERP evidence?

Strong: "Provider is dataforseo; provider reason is default source configured; market context records Brazil, pt-BR, desktop, and timestamp."

Weak: "Provider is websearch because it was faster, with no bypass record."

If DataForSEO cannot be used, stop before analysis and record a bypass reason. The bypass record must state that WebSearch is a fallback, the result may miss metrics or exact SERP ordering, and the artifact will disclose the bypass. A bypass record is not evidence for missing metrics.

3. Gather And Normalize Evidence

Check: Are extracted facts stored separately from interpretation?

Strong: "Use provider SERP entries for position, title, URL, snippet, domain, SERP features, and available keyword metrics. Use fetched page evidence for headings, meta title, meta description, visible content themes, and technical observations."

Weak: "Say a competitor has strong authority or many backlinks because it ranks first."

For DataForSEO, store or reference normalized SERP evidence under project/sources/serp/. For WebSearch fallback, store or reference results under project/sources/websearch/ and set keyword metrics to null unless another source provides them.

4. Compare The Top 3

Check: What do the top 3 pages reveal about intent, page type, proof, structure, and missing angles?

Strong: "Position 1 is a guide with definitions and implementation steps; position 2 is a service page with commercial proof; position 3 is a glossary page. The likely mixed intent is informational with commercial evaluation."

Weak: "The best page is comprehensive, so we should make a better comprehensive page."

Ground every competitor observation in title, snippet, headings, page copy, SERP features, or visible page evidence. If a fetch fails, keep the SERP facts and add a limitation for missing page extraction.

5. Interpret Target Gaps And Player Score

Check: How does the target page compare with ranking pages, and what does the score mean?

Strong: "The target is outside the top 10. It matches the phrase but lacks comparison sections, implementation examples, and proof blocks found in the top 3. Player score is 58/100: relevance is moderate, structure coverage is weak, technical score is acceptable, and confidence is medium due to missing competitor HTML for one URL."

Weak: "The target should rank after adding more keywords."

When scoring, separate deterministic observations from judgment. Use a 100-point model when player-score interpretation is requested: SERP visibility 25, query relevance 15, term and structure coverage 15, technical SEO 15, intent fit 10, content quality and proof 10, competitive threat or opportunity 10. Include confidence separately from score.

6. Produce Hypotheses And Next Actions

Check: Which improvements are evidence-backed, and which remain hypotheses?

Strong: "Hypothesis: adding a section on SEO agêntico vs automação de SEO may improve intent fit because two top results address automation boundaries."

Weak: "This will increase traffic by 40%."

Write hypotheses as testable ideas, not promises. Include evidence references and limitations so a human can decide whether to use them or request more research.

Output Format

Write the report to project/workbench/seo-analysis/<keyword-slug>.yaml unless the user asks for an inline preview first. Use this structure:

status: complete | blocked | incomplete
keyword: ""
provider: dataforseo | websearch
provider_reason: ""
websearch_bypass:
  recorded: true | false
  registrado_por: agent
  note: null
  reason: null
  consequence: null
  timestamp: null
market_context:
  country_or_market: ""
  location: ""
  language: ""
  device: desktop | mobile
  generated_at: ""
keyword_metrics:
  volume: null
  cpc: null
  competition: null
sources:
  serp:
    - path: project/sources/serp/...
  websearch:
    - path: project/sources/websearch/...
  pages:
    - url: ""
      evidence_ref: ""
top_results:
  - position: 1
    title: ""
    url: ""
    domain: ""
    snippet: ""
    observed_headings: []
    observed_meta: {}
top_3_comparison:
  intent_pattern: ""
  page_type_pattern: ""
  structure_pattern: ""
  proof_pattern: ""
  missing_or_weak_evidence: []
target:
  mode: url | domain | none
  url: null
  domain: null
  ranking_status: exact_url_ranking | same_domain_wrong_url | domain_ranking | not_ranking | unknown
  gaps: []
player_score:
  score: null
  confidence: null
  components: []
  interpretation: null
synthesis:
  intent: ""
  opportunities: []
  risks: []
hypotheses:
  - hypothesis: ""
    evidence_refs: []
    validation_needed: ""
limitations: []
next_actions: []

If blocked by a missing provider or missing bypass record, return status: blocked, describe the gate, and do not invent a partial SERP.

Default delivery

Follow the shared page-report contract and the module skeleton at templates/analyses/seo-analysis/report-skeleton.md. The module-specific source artifact is the normalized SEO-analysis YAML under audits/seo-analysis-<keyword-slug>/report.yaml or workbench/seo-analysis/; the Companion page is project/analyses/seo-analysis/<keyword-slug>/report.md. Start with the executive reading and priorities, then show competitor/player details and limitations as readable tables and prose. Translate labels/status/severities/nulls, use friendly check names, and never paste raw SERP, heading, player-score, or provider JSON into the visual report body. The response must return report_md plus browser_prompt: { recommended: true, message: "Posso abrir o Web Companion para você ver a análise?", open_with: "project-browser" }.

Examples

Example: DataForSEO Analysis

Input: "Analyze seo agêntico for Brazil in pt-BR desktop. Target page is https://example.com/seo-agentico/."

Output: "Use DataForSEO, record Brazil, pt-BR, desktop, timestamp, compare the top 3 organic pages, mark the target as outside top 10 if supported by evidence, interpret page gaps and player score, and write hypotheses with accents preserved in Portuguese prose."

Example: WebSearch Bypass

Input: "DataForSEO is unavailable. Use WebSearch anyway because this is only a rough exploratory analysis."

Output: "Use provider: websearch, record the bypass reason and consequence, set unavailable keyword metrics to null, disclose limitations, and keep ranking claims limited to the captured WebSearch results."

Example: Weak Execution

Input: "Analyze the ranking opportunity for seo agêntico."

Output: "Search the web, guess that volume is high, say competitors have strong backlinks, and recommend publishing to the brain." This is weak because it bypasses DataForSEO without a record, fabricates metrics and backlinks, and treats hypotheses as decided strategy.

Related Skills

• keyword-research: use when the primary task is keyword discovery, clustering, or metric collection before SERP analysis.
• content-seo: use after this analysis when the user wants a content brief or draft.
• topic-cluster: use after enough evidence-backed analysis exists to organize topics into a cluster.
• technical-seo: use when the primary task is crawling, rendering, or auditing page health rather than interpreting one keyword SERP.