SEO Local
by agricidaniel / claude-seo
Audit local SEO, including Google Business Profile, NAP consistency, citations, reviews, local schema, location pages, and multi-location structure.
Extracts each location's name, address and phone from your site and major directories, then flags inconsistencies, formatting drift and gaps.
Listed ·Updated
AGENT ROLE: Autonomous NAP-integrity agent. Extract each location's Name/Address/Phone from every source, normalize, diff against the canonical record, and emit the JSON in references/output.schema.json.
For each business location, establish a canonical NAP and detect every place (own site pages, citation directories) where it differs — even by formatting — plus flag directories where the location is missing.
locations (REQUIRED): array of { id, canonical: { name, address, phone }, page_urls: string[] }.citation_urls (OPTIONAL): map of location_id → [directory listing URLs] to check (Yelp, BBB, Apple Maps, industry directories).strict_phone (OPTIONAL bool, default false): if true, formatting differences in phone count as mismatches; default normalizes to digits.seoskills-nap/1.0, honor robots. Some directories block bots — treat blocks as unverifiable, not mismatch.scripts/nap_crawl.py --locations locations.json [--citations citations.json].LocalBusiness/PostalAddress JSON-LD first, then microdata, then visible-text heuristics.STEP 1 — CANONICAL: normalize the supplied canonical NAP (see normalization below).
STEP 2 — For each page_url and citation_url: fetch, extract NAP (schema → microdata → regex fallback), normalize.
STEP 3 — NORMALIZE: name → lowercase, strip legal suffixes/punctuation; address → USPS-style abbreviations (Street→St, Suite→Ste), collapse whitespace; phone → digits only (unless strict_phone).
STEP 4 — DIFF each source's normalized NAP vs canonical; classify per field match | mismatch | missing. A source with any field mismatch = inconsistent.
STEP 5 — MISSING CITATIONS: any citation_url that returns no detectable listing for the location → missing_citation.
STEP 6 — SCORE consistency = matched_fields / total_checked_fields per location; EMIT inconsistencies grouped by location, each with the exact source URL and the field-level before/after.
403/429/CAPTCHA THEN mark that source unverifiable (NOT a mismatch) and continue — never assert inconsistency from a blocked fetch.5xx ≤ 2.extraction="failed" for that source (distinct from a real mismatch).normalized_match and the raw values.One JSON object per references/output.schema.json. No prose.
scripts/nap_crawl.py — multi-source NAP extraction, normalization, diff, scoring.references/output.schema.json — output contract.Not using the CLI? Copy the SKILL.md and paste it straight into ChatGPT, Claude, or any agent.