Healthcare Providers Extract

• You have practice URLs and need provider data. Paste one URL or a list of 100+; the skill maps each site, finds provider pages, and returns structured records.
• You want to build or refresh a provider database for a specific market. Give it a specialty and location (e.g., "ophthalmologists in Austin, TX") and it will discover practices via Google Maps before extracting.
• You need batch extraction across many practices. The sub-agent strategy parallelizes site mapping and extraction, making it practical for large research or sales prospecting jobs.
• You require source-traceable data. Every provider record links back to its source URL, making the output audit-ready.

• You need to fill gaps in an existing provider dataset — use healthcare-providers-enrich instead.
• You need to validate credentials or license status — use healthcare-providers-verify instead.
• You need to discover practices (not extract from them) — use market-finder or local-places instead.
• You need general-purpose web extraction unrelated to healthcare providers — use nimble-web-expert instead.

For every provider found on a practice site, the skill extracts up to five fields: full name, credentials (e.g., MD, FACS, OD), specialty, contact information (phone, scheduling URL, email), and education/training mentions. A confidence score (High / Medium / Low) flags how complete each record is.

Accepts practice website URLs pasted directly, uploaded as a CSV, or linked from a Google Sheet. If you don't have URLs, provide a specialty and location and the skill will discover matching practices via Google Maps before extracting.

For jobs covering six or more practices, the skill spawns parallel nimble-researcher sub-agents — one per practice or per batch of three — so large extractions complete without blocking. Single-practice runs execute directly from the main context for speed.

Before extracting, the skill maps each practice domain to find the right pages (team, providers, doctors, staff) using a healthcare-specific keyword weight table, capped at 15 pages per site. This avoids wasting extraction calls on irrelevant pages.

Provider records are deduplicated across pages and practices, then scored for confidence based on how many of the five fields were populated. The output groups providers by practice and sorts by confidence, with a Data Quality Summary highlighting common gaps.

The skill tracks previous runs and applies smart date windowing: full extraction on first run or after 14+ days, quick refresh (new/changed pages only) within 14 days, and a confirmation prompt if re-run the same day. Results are saved to memory for follow-up actions like CSV export, enrichment, or distribution.

A medical device sales team provides a CSV of 80 ophthalmology practice URLs across three states. The skill extracts every provider's name, credentials, specialty, and phone number, returning a table sorted by confidence that feeds directly into a CRM.

A healthcare analytics firm specifies "dermatologists in Miami, FL." The skill discovers practices via Google Maps, resolves their website URLs, and extracts structured provider profiles across all discovered sites — producing a market snapshot without any manual URL gathering.

A health insurance company needs to populate a new in-network directory. They paste URLs for 120 contracted practices; the skill extracts all providers with education and credential data, flags low-confidence records for manual review, and exports the result as a CSV.

A multi-location dental group wants to understand competitor staffing. They provide URLs for ten rival practices; the skill maps each site, extracts provider profiles, and surfaces credential and specialty breakdowns grouped by practice.