SEO Content
by agricidaniel / claude-seo
Evaluate page content for usefulness, E-E-A-T, readability, thinness, and AI citation readiness, plus last-mile draft cleanup (AI-typical phrasing and invisible Unicode watermark characters).
Scores a page's E-E-A-T signals, such as author credentials, citations, first-hand experience, freshness and outside corroboration.
Listed ·Updated
AGENT ROLE: Autonomous trust-signal agent. Detect and score each E-E-A-T signal on a page against the rubric in references/eeat_rubric.json and emit the JSON in references/output.schema.json. Score observable signals only; never assert authority the page does not demonstrate.
Produce a per-signal E-E-A-T scorecard (Experience, Expertise, Authoritativeness, Trust) with the specific missing elements to add — grounded in what is actually present on the page and its author entity.
url (REQUIRED): the page to evaluate.topic (OPTIONAL): the page's subject, to check author-topic relevance.is_ymyl (OPTIONAL bool): if true, apply the rubric's stricter YMYL weights.verify_author_entity (OPTIONAL bool, default true): look the author up in the Knowledge Graph / Wikidata.seoskills-eeat/1.0, honor robots.verify_author_entity: uses the Google Knowledge Graph API IF KG_API_KEY is set, else Wikidata (keyless). IF neither resolves the author THEN the authoritativeness sub-check is marked unverifiable, not failed.scripts/eeat.py --url {url} [--topic "..."] [--ymyl].Article/Person schema, author, datePublished/dateModified), byline, bio, outbound citations, and experience-language cues.STEP 1 — FETCH + PARSE: extract author name (schema author → byline → rel=author), author bio/credential text, datePublished/dateModified, outbound links to authoritative domains, and citation/reference markers.
STEP 2 — EXPERIENCE: detect first-hand markers (references/eeat_rubric.json cue lists: "I tested", "we measured", original images, "in my experience"). Score by cue density + original-media presence.
STEP 3 — EXPERTISE: author bio contains credentials/qualifications relevant to topic; presence of a detailed author bio and Person schema with jobTitle/knowsAbout.
STEP 4 — AUTHORITATIVENESS: resolve the author entity (KG/Wikidata) and count authoritative outbound citations; site-level sameAs/organization signals.
STEP 5 — TRUST: dateModified recency, HTTPS, contact/about presence, citation of sources, absence of deceptive patterns, Article schema completeness.
STEP 6 — SCORE each pillar 0–100 via the rubric weights (YMYL weights if is_ymyl); overall = weighted mean. EMIT per-pillar scores + the concrete missing_signals list.
error.code="PAGE_UNREACHABLE" with the status.429 → backoff 2^attempt (max 4); on failure set authoritativeness author-check unverifiable.error.code="ROBOTS_DISALLOW" unless respect_robots=false.missing_signals: ["author_identity"]), not an error.One JSON object per references/output.schema.json. No prose.
scripts/eeat.py — page parse, signal extraction, entity verification, rubric scoring.references/eeat_rubric.json — signal cues, weights, and YMYL overrides.references/output.schema.json — output contract.Not using the CLI? Copy the SKILL.md and paste it straight into ChatGPT, Claude, or any agent.