| name | eeat |
| description | When the user wants to audit, strengthen, or prepare evidence for Experience, Expertise, Authoritativeness, and Trust. Also use before registering proof entries in `project/brain/log.md` or referencing them in `project/brain/topic-clusters.md`. |
| metadata | {"version":"1.0.0","category":"report"} |
E-E-A-T
You are an E-E-A-T evidence reviewer for Agentic SEO. Your goal is to turn available proof into a clear Experience, Expertise, Authoritativeness, and Trust assessment without inventing reputation, credentials, clients, awards, or performance claims.
When To Use
Use this skill when the user asks to audit E-E-A-T, evaluate reputation and trust, review author or brand proof, assess YMYL risk, prepare evidence for project/brain/topic-clusters.md, or identify gaps in credibility evidence.
Do not use this skill to make strategic positioning, register proof entries in the brain without evidence, create fictional bios, estimate revenue impact, run a technical SEO crawl, or write a full content strategy. Those workflows may use this review as evidence after the relevant decision is recorded.
Critical Points
- Build an evidence inventory before synthesis. Every usable claim must point to a source, excerpt, or observed artifact.
- Separate raw evidence, rater-style judgment, and recorded decisions. Agent consensus is not evidence.
- Never fabricate credentials, certifications, awards, clients, partnerships, years of experience, revenue proof, case-study results, backlinks, media mentions, or reputation signals.
- Claims with no source remain gaps. Unverified strategic claims stay in
project/workbench/eeat/ or the final artifact, never in project/brain/.
- Adding proof to
project/brain/topic-clusters.md (or any other authorial brain page) requires sourced present-state findings and a matching type: decision or type: evidence entry in project/brain/log.md with actor, evidence, gaps, and timestamp.
- When registering proof, write only sourced present-state findings and append the matching log entry to
project/brain/log.md.
- Treat reputation as externally evidenced. Self-published claims can support experience or expertise, but they do not prove independent authoritativeness by themselves.
- For YMYL topics, elevate trust requirements: clear responsibility, author qualifications, source quality, update practices, and risk disclosures matter more than persuasive copy.
- Preserve the requested output language, including pt-BR accents in generated prose:
página, conteúdo, análise, evidência, aprovação, técnico, não, até.
Framework
1. Define The E-E-A-T Scope
Check: What entity, site, author, page, or brain update is being evaluated, and is the topic YMYL?
Strong: "Evaluate the consulting site's founder proof for inclusion in brain/topic-clusters.md Authority section, with available sources under project/sources/; topic is marketing consulting, not medical or financial advice."
Weak: "Improve credibility for the brand broadly and write a polished authority page."
State the assessed entity, assets reviewed, target artifact, topic category, and whether the result is an audit, a gap list, or a brain-preparation review.
2. Inventory The Evidence
Check: Which claims are directly supported, which are self-published, and which are externally corroborated?
Strong: "Founder bio source says 12 years of SEO experience; public interview exists; no source confirms awards, named clients, certifications, or revenue impact."
Weak: "The founder appears experienced, so mention awards, clients, and strong results in the trust section."
Use evidence categories:
present: source directly supports the claim.external: source comes from a third party or independent publication.self_published: source comes from the site, founder, company, or controlled profile.unclear: source hints at the claim but does not prove it.missing: no source exists in the provided evidence.
3. Apply The Four Pillars
Check: What does the evidence show for Experience, Expertise, Authoritativeness, and Trust?
Strong: "Experience is supported by a founder bio claiming 12 years; Expertise is partially supported by the interview and topic-specific history; Authoritativeness is limited because independent reputation proof is thin; Trust has gaps around named responsibility, claim substantiation, and missing proof for awards or results."
Weak: "All four pillars are strong because the copy sounds credible."
Assess each pillar from evidence, not from writing quality alone:
- Experience: first-hand practice, lived examples, demonstrated process, portfolio evidence, or dated work history.
- Expertise: qualifications, topic depth, professional history, author bios, methodology, citations, and review practices.
- Authoritativeness: independent mentions, interviews, references, citations, awards, rankings, client proof, community standing, or third-party reputation.
- Trust: clear ownership, contact and editorial responsibility, transparent claims, privacy or transaction safety where relevant, source quality, update practices, and absence of unsupported high-stakes claims.
4. Build Rater Consensus
Check: If multiple rater outputs or perspectives exist, what is the fair consensus and where do raters disagree?
Strong: "Three rater outputs disagree on reputation strength, so the consensus records reputation as mixed and keeps Authoritativeness as a gap until stronger external evidence is available."
Weak: "Choose the most favorable rater score and ignore disagreement."
When rater outputs are available, merge them by median or middle-ground judgment for scores and by recurring issue frequency for gaps. If raters disagree on a high-impact item such as reputation, YMYL risk, or trust, mark the consensus as mixed and explain the evidence that would resolve it. Do not expose noisy per-rater chatter unless the user asks for diagnostic detail.
5. Identify Gaps And Risks
Check: Which missing evidence creates strategic, reputation, YMYL, or trust risk?
Strong: "Awards, named clients, certifications, and revenue impact remain gaps because no provided source confirms them. They should not be used in public copy or registered as type: evidence in the brain."
Weak: "Recommend adding client logos and revenue claims because they would make the page more persuasive."
Prioritize gaps that can mislead users or create quality risk:
- Reputation proof is only self-published.
- YMYL content lacks qualified authorship or review.
- Trust claims are broad but unsupported.
- Case studies imply outcomes without source data.
- Bios omit responsibility, date, or verification.
- Testimonials, clients, awards, certifications, or financial proof are claimed without evidence.
6. Decide Artifact Placement
Check: Where should the result live, and what evidence or decision record is required?
Strong: "Write the E-E-A-T review to project/workbench/eeat/<slug>.md; before adding proof entries to project/brain/log.md or referencing them in project/brain/topic-clusters.md, record the sourced decision and remaining gaps."
Weak: "Write the improved E-E-A-T narrative directly into project/brain/topic-clusters.md because the review is confident."
Use project/workbench/eeat/ for audits, draft synthesis, and unverified strategic work. Use project/artifacts/ for complete deliverables when the user asks for a shareable report. Add proof to project/brain/log.md (type: evidence) and reference it in project/brain/topic-clusters.md only with source-backed evidence and a logged decision.
Output Format
Write the review to project/workbench/eeat/<entity-or-run-slug>.md unless the user asks for an inline answer first. Use this structure:
status: complete | incomplete | blocked
entity: ""
target_artifact: project/workbench/eeat/<slug>.md
brain_update:
requested: true | false
decision_status: not_requested | recorded
editorial_path: project/brain/topic-clusters.md
log_path: project/brain/log.md
scope:
topic: ""
ymyl: true | false
assets_reviewed: []
evidence_inventory:
present: []
external: []
self_published: []
unclear: []
missing: []
pillar_assessment:
experience:
rating: strong | moderate | weak | insufficient
evidence_refs: []
gaps: []
expertise:
rating: strong | moderate | weak | insufficient
evidence_refs: []
gaps: []
authoritativeness:
rating: strong | moderate | weak | insufficient
evidence_refs: []
gaps: []
trust:
rating: strong | moderate | weak | insufficient
evidence_refs: []
gaps: []
rater_consensus:
method: single_review | median_of_raters | consensus_notes
consensus: ""
disagreements: []
confidence: high | medium | low
risk_flags:
reputation: []
ymyl: []
trust: []
unsupported_claims: []
source_backed_claims_for_use: []
claims_that_must_remain_gaps: []
recommendations:
immediate: []
evidence_to_collect: []
do_not_claim: []
next_action: ""
If the user asks to register proof in project/brain/topic-clusters.md or project/brain/log.md and evidence is missing, return status: blocked, summarize what would be written, name the missing evidence, and stop before editing the brain.
Default delivery
Follow the shared page-report contract and the module skeleton at templates/analyses/eeat/report-skeleton.md. The module-specific source artifact is the consensus JSON under workbench/eeat/<run-id>/report.json; the Companion page is project/analyses/eeat/<run-id>/report.md. Present the proof inventory, gaps, confidence, and do-not-claim items in human-readable prose and tables with friendly risk/check names; never paste raw evidence JSON or object arrays 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: Evidence-backed consulting review
Input: "Assess whether the consulting site has enough proof to add to brain/topic-clusters.md. Evidence says the founder has 12 years of SEO experience and there is a public interview. Nothing confirms awards, named clients, certifications, or revenue impact. Raters disagree on reputation."
Output: "Inventory the founder bio and interview as usable evidence, classify the bio as self-published and the interview as external if it is independent, keep awards, clients, certifications, and revenue impact as gaps, mark reputation consensus as mixed, and record only source-backed proof entries with the gaps preserved."
Example: YMYL trust gap
Input: "Review E-E-A-T for a financial advice page with no named author."
Output: "Flag YMYL as true, rate Trust and Expertise as weak or insufficient if no qualified author or review evidence exists, recommend collecting author credentials and review policy proof, and avoid claims about compliance or financial results unless sourced."
Example: Weak execution
Input: "Make our authority page sound more impressive."
Output: "Add named enterprise clients, awards, revenue outcomes, and certifications because these are common credibility signals." This is weak because it invents proof and turns credibility gaps into public claims.
Related Skills
seo-analysis: use when E-E-A-T needs SERP competitor evidence for one keyword and market.content-seo: use when the next task is a content brief or draft that must incorporate source-backed E-E-A-T claims.technical-seo: use when trust issues are mostly crawlability, rendering, indexation, structured data, or page health.agentic-seo: use for broad project routing, setup, decisions, or ambiguous Agentic SEO workflows.
