| name | seo-eeat-ymyl-google |
| description | Apply when a site-wide audit according to Google Search Quality Rater Guidelines is needed with a focus on E-E-A-T, YMYL, Page Quality, and Needs Met for medical, financial/transactional, and standard content sites with a scorecard-first output. |
SEO E-E-A-T YMYL Google
Using seos-cli in this skill
Adoption decision: seos-cli is optional and not used by default. This skill's core work is QRG judgment, hard-gate analysis, E-E-A-T scoring, and evidence-matrix synthesis; seos-cli does not perform that reasoning.
Use seos-cli only as supplemental evidence collection when project-native page text or SERP evidence would materially help the audit:
seos-cli config check
seos-cli scrape --urls https://example.com/page --format json --out outputs/eeat-page-scrape.json
seos-cli serp --keywords "brand query, product query" --format xlsx --out outputs/eeat-serp-evidence.xlsx
Do not use seos-cli when local HTML, WebFetch, a subagent scrape, reputation research, or scripts/extract_eeat_signals.py already provides the evidence needed by this skill. Do not claim seos-cli performs E-E-A-T scoring, YMYL classification, Page Quality, Needs Met, or final QRG verdicts.
Overview
QRG-first skill for site-wide quality, trust, and risk audit according to Google Search Quality Rater Guidelines 2025.
Core principle: hard-gates and evidence first, then Page Quality, then E-E-A-T scorecard. Needs Met is evaluated only if a query context is available.
When to apply
- A site audit or URL sample audit is needed for
E-E-A-T, YMYL, Page Quality, trust, reputation, harm.
- Need to separate low-trust YMYL/transaction pages from standard informational pages.
- Need to generate a
scorecard-first report, not a narrative without metrics.
- Need to quickly and repeatedly extract on-page signals before manual judgment.
Do not use as a crawl-heavy technical audit. For leak/proxy models, graph scoring, and massive crawling, use other skills.
Discovery scenarios
- Medical site: check YMYL-risks, trust-signals, author, and review process.
- Financial/transactional site: check safety, transaction transparency, and customer-service trust signals.
- Standard content site without obvious YMYL: check usefulness, content quality, and basic trust.
Expected input
- Domain or list of URLs.
- If available: query or query-cluster for
Needs Met.
- If necessary: local HTML files for selective checking.
Workflow
- Determine the site type and
YMYL Tier: Non-YMYL | Possible YMYL | Clear YMYL.
- Collect a sample of 5 URLs: homepage, money/service/product, informational, trust/about/contact/author, highest-risk YMYL or transaction.
- For webfetch, scraping, external reputation search, and mass page reading, invoke a subagent first.
- Pass hard-gates: harmful purpose, deceptive purpose/design, scaled low-value content, reputation abuse, hidden/blocked main content, unsafe sensitive-data requests, untrustworthy YMYL claims.
- If necessary, run
scripts/extract_eeat_signals.py to get a compact JSON for the pages.
- Assign
Page Quality: Lowest | Low | Medium | High | Highest.
- Fill out the E-E-A-T scorecard:
Experience, Expertise, Authoritativeness, Trustworthiness, Overall E-E-A-T.
- If there is a query context, evaluate
Needs Met separately.
- Generate the final report according to
references/output-contract.md. If desired, use scripts/render_scorecard.py only to prefill the template, not for final judgment.
Mandatory rules:
- Do not mix
Page Quality and Needs Met.
- If query context is missing, set
N/A for Needs Met.
- hard-gate findings can forcefully lower the result regardless of the formal score.
- Explicitly distinguish
Experience and Expertise; in high-risk YMYL, do not substitute formal expertise with personal experience.
scripts/render_scorecard.py can be used only for prefill and structure normalization, but not for automatic verdicting.
What to read
references/qrg-page-quality.md: hard-gates and Page Quality logic.references/ymyl-thresholds.md: YMYL Tier, trust requirements, experience vs expertise.references/eeat-scorecard-model.md: rubric for E-E-A-T scoring.references/needs-met-and-intent.md: when and how to use Needs Met.references/qrg-metric-catalog.md: full metric register that must be included in the report.references/output-contract.md: mandatory sections, evidence format, and wording rules.
Lite Scripts
scripts/extract_eeat_signals.py
- Extracts on-page signals from URLs or local HTML.
- Returns a compact JSON for scorecard prep.
scripts/render_scorecard.py
- Renders the markdown report template and evidence matrix.
- Does not make final decisions instead of the agent.
Language and context rules
- Write the main text of the skill, reference files, templates, and report in English.
- Use English terms only as standard labels:
E-E-A-T, YMYL, Page Quality, Needs Met, Lowest, Highest, Person schema.
- Do not create bilingual duplicates.
- Delegate long scraping/webfetch/reputation-heavy stages to a subagent; the main agent receives only a compressed summary, extracted signals, and findings.
- Keep decision-making, scorecard assembly, and final report synthesis local.
