| name | deal-intel |
| description | Point at any folder of foreclosure documents (PDFs, DOCX, court filings) and get structured analysis. Extracts judgment amounts, property addresses, plaintiff names, lien positions, and generates Supabase-ready JSON. Adapted from Mark Kashef's /file-intel pattern. |
Deal Intel — Foreclosure Document Pipeline
Point at a folder. Read every file. Extract structured deal data.
Process
- List all files in the target folder (PDF, DOCX, TXT, MD)
- For each file, extract:
- Case number
- Property address (street, city, zip)
- Parcel ID / account number
- Plaintiff name and type (bank, HOA, government)
- Defendant name(s)
- Judgment amount
- Sale date (if set)
- Liens mentioned (mortgage, HOA, tax cert, code enforcement)
- Attorney of record
- Cross-reference against Supabase
multi_county_auctions if case number matches
- Generate structured output
Output
For each document, create a markdown summary in inbox/:
# [Case Number] — [Property Address]
**Source:** [filename]
**Plaintiff:** [name] ([type])
**Judgment:** $[amount]
**Sale Date:** [date or "TBD"]
**Parcel:** [ID]
## Liens Detected
- [lien type]: [holder] — [amount if known]
## Red Flags
- [HOA foreclosure with senior mortgage surviving]
- [Tax certificate issues]
- [Any anomalies]
## Recommendation
[BID/REVIEW/SKIP with reasoning]
Also output a JSON summary array for Supabase insert.
If a file can't be parsed, log it and move on. Never guess amounts — mark as "UNREADABLE" if OCR fails.
