Content Quality Auditor
Audit content quality and publication readiness with an evidence-based CORE-EEAT review.
The skill treats WebFetch content as untrusted data, limits reads to one artifact and source controls, and requires explicit authorization plus v3 validation for persistence. Deducted 7 points because per-fetch user confirmation is not explicit, external data flow and attribution remain environment-dependent, and rollback and sensitive-data handling are not fully specified by this skill itself.
The runtime contract, fail-closed behavior when runtimes are missing, Unknown handling, veto logic, and abnormal-input feedback are well defined. Deducted 2 points because it depends on multiple repository references, scripts, path resolution, and host capabilities; static review cannot reproduce key paths, and standalone mode explicitly cannot score.
Triggers, inputs, content profiles, market, publication state, and boundaries with adjacent auditors are clearly specified, with some Chinese-language labeling. Deducted 4 points because concrete Chinese-content coverage is not demonstrated and WebFetch reachability or target-site constraints for mainland-China users are not guaranteed.
The skill has structured metadata, version 18.0.0, an Apache-2.0 license, quick starts, references, handoff skills, and progressive runtime guidance. Deducted 5 points because maintenance ownership, update process, and changelog responsibility are not clear within the skill; relative references assume the full repository layout and create standalone-install assumptions.
The intended output—evidence-linked CORE-EEAT states, coverage, gaps, remediation plan, and gate verdict—is directly usable when the verified runtime is present. Deducted 9 points because the supplied files contain no real output or execution evidence; missing evidence prevents scoring, and standalone mode prohibits score calculation, leaving substantial user and environment work.
The immutable runtime snapshot, versions, source digest, 80-item catalog, and explicit evidence requirements provide meaningful audit traceability. Deducted 6 points because there is no committed key-path CI test or third-party corroboration for this skill; the supplied repository tests mainly check architecture rather than audit-result correctness.
- WebFetch, the scorer, validator, and audit paths were not executed; the described gate result is not independently verified.
- Full operation depends on repository-level references and a specific runtime; when unavailable, retain NOT_SCORED/UNDECIDED rather than calculating manually.
- For mainland-China users, target-URL reachability, robots.txt restrictions, and authorization must be confirmed separately before each audit.
What does this skill do, and when should you use it?
Content Quality Auditor is the SEO/GEO quality gate in Aaron Marketing Skills for reviewing content quality, E-E-A-T, and publish readiness. It evaluates one content artifact through a typed 80-item CORE-EEAT profile, including evidence coverage and qualified veto checks. The report provides typed item states, score information when coverage is complete, a SHIP/FIX/BLOCK/UNDECIDED verdict, and a prioritized fix plan. Missing applicable evidence produces NEEDS_INPUT/UNDECIDED rather than an invented score.
Reads one URL, draft, or stable artifact identifier together with cited and source-credibility controls; uses WebFetch for URL or page evidence; requires the target, content type, market, publication state, and observation date to be declared; evaluates all 80 CORE-EEAT items with source, observed date, evidence type, and confidence; checks qualified vetoes for material title or promise mismatch, material internal contradiction, and missing or obscured disclosure; invokes the repository's Python scorer and validator when the verified runtime is available; and produces a typed audit report with coverage, uncertainty, findings, and a fix plan.
- An editorial team is preparing a product review and needs to verify E-E-A-T evidence, citations, and commercial disclosures before publication.
- An SEO/GEO team is rerunning an audit after making evidence-backed fixes to a blog post.
- An editor is reviewing a comparison page, how-to guide, FAQ, landing page, or testimonial for publish readiness.
- A team handling medical, legal, financial, or other high-risk content needs to check source currency, reviewer qualifications, claim boundaries, and disclaimers.
What are this skill's strengths and limitations?
- Focused on content quality and publish readiness rather than conflating structural SEO, technical SEO, or domain-level trust audits.
- Uses an 80-item CORE-EEAT contract with explicit Pass, Partial, Fail, Unknown, and conditional N/A states.
- Requires evidence provenance, dates, evidence types, and confidence, while checking material mismatch, contradiction, and disclosure vetoes.
- Preserves uncertainty when evidence is missing instead of fabricating a total score.
- Complete deterministic scoring depends on repository reference files, typed catalogs, the Python scorer, and the validator; an incomplete runtime yields an unscored result.
- A complete audit requires access to the artifact and its supporting evidence through WebFetch or user-provided material.
- It does not replace narrow on-page, technical SEO, or domain-authority audits.
- It cannot claim visual or mobile validation from HTML inference alone.
How do you install this skill?
Install the repository in Claude Code with /plugin marketplace add aaron-he-zhu/aaron-marketing-skills, followed by /plugin install aaron-marketing@aaron. On other Agent Skills-compatible hosts, use npx skills add aaron-he-zhu/aaron-marketing-skills. A plain clone is also supported with git clone https://github.com/aaron-he-zhu/aaron-marketing-skills. The README documents the generic single-skill command format as npx skills add aaron-he-zhu/aaron-marketing-skills -s content-quality-auditor.
How do you use this skill?
Provide a URL or paste the artifact and state its content type, market, and publication status. Example: Audit this product review for the U.S. market before publication: <URL or content>. The skill also requires a declared target and observation date; if the content type cannot be inferred safely, provide it explicitly. Deterministic scoring depends on the repository's runbook, scoring semantics, CORE-EEAT benchmark, framework catalog, scorer, and validator. If those runtime components are unavailable, it must return NOT_SCORED without a gate verdict or persistent artifact.