`PROMPT.txt`

91 lines
# Claude Code Master Prompt — LABJ CRE + Banking Public-Source Dataset

Build and maintain a research dataset of Los Angeles Business Journal (LABJ) “The Lists” companies relevant to commercial real estate, banking, lending, construction, property management, architecture, and potential RENTV advertisers.

STARTING FILE:
labj_cre_banks_public_seed.csv

IMPORTANT LEGAL / ACCESS RULES:
1. Use only information that is publicly accessible without bypassing authentication, subscription controls, CAPTCHAs, robots restrictions, or technical access controls.
2. Do not attempt to download or reconstruct LABJ paid Excel/PDF products unless the user supplies a legitimately obtained copy.
3. LABJ states its lists may not be reprinted in whole or in part without permission. Treat LABJ as a discovery/ranking source; enrich records from independent public sources and keep source attribution.
4. Respect website terms, robots.txt, reasonable request rates, and copyright.
5. Never fabricate missing fields.

TARGET LABJ CATEGORIES:
- Commercial Real Estate Brokerage Firms
- Commercial Real Estate Developers
- Property Management Firms
- Property Management Firms — Office
- Property Management Firms — Retail
- Property Management Firms — Industrial
- General Contractors
- Architecture Firms
- Construction Projects
- Infrastructure Construction Projects
- Residential Real Estate Developers
- Residential Brokerage Firms
- Banks
- Financial Institutions
- Credit Unions
- SBA 7(a) Lenders
- SBA 504 Lenders
- Private Equity Firms
- Venture Capital Firms
- Securities Brokerage Firms

WORKFLOW:
A. Inspect the seed CSV and preserve all existing rows and source URLs.
B. Discover current LABJ public HTML list/product pages and record list name, publication year/date, ranking basis, public rows visible, and LABJ URL.
C. For each discovered company, enrich from the company’s official website and other lawful public sources:
   company_name, aliases, category, LABJ_rank, LABJ_list, LABJ_year,
   address, city, state, zip, website, main_phone,
   top_local_executive, title,
   marketing_executive, marketing_title,
   PR_contact, PR_title, public_email,
   LinkedIn_company_url, employee_count_if_public,
   specialties, markets, LA_presence,
   advertiser_fit_score, advertiser_fit_reason,
   source_labj, source_company, source_contact, last_verified.
D. Deduplicate by normalized company/domain while preserving multiple LABJ list memberships.
E. Score advertiser fit 0–100 for RENTV:
   +25 commercial real estate core business
   +20 active Los Angeles presence
   +15 marketing/PR contact identified
   +15 appears on multiple relevant lists
   +10 lender/title/escrow/bank serving CRE
   +10 major transaction/development/property-management footprint
   +5 public advertising/sponsorship evidence
F. Output:
   1. data/labj_master.csv
   2. data/labj_companies.csv
   3. data/labj_contacts.csv
   4. data/labj_list_memberships.csv
   5. data/research_queue.csv
   6. reports/top_100_rentv_prospects.csv
   7. reports/source_audit.csv
   8. README.md describing sources, limitations, and refresh process.

TECHNICAL:
- Python 3.
- Prefer requests + BeautifulSoup for public static pages.
- Use Playwright only where normal public page rendering requires JavaScript.
- Add a descriptive User-Agent.
- Rate limit requests.
- Cache fetched pages.
- Log URL/status/timestamp.
- Do not evade 401/403/429 responses.
- Do not use credentials unless explicitly supplied by the user for authorized access.
- Validate URLs and normalize domains.
- Keep raw source attribution for every researched field where practical.

FIRST RUN:
1. Load labj_cre_banks_public_seed.csv.
2. Create the folder structure.
3. Research and enrich the existing companies first.
4. Discover additional relevant public LABJ list pages.
5. Expand with independently verified public company/contact data.
6. Print a completion summary: total companies, contacts, categories, verified URLs, unresolved records, and top 25 RENTV prospects.

Execute the build now. Do not stop at planning; create the files and populate them with all lawfully accessible public data you can verify.