--- name: owner-mailing-list description: Produce a deduplicated owner + mailing-address list for a Lee & Associates broker from county parcel data. Given an area + criteria request (subject address + radius + property type/land class + acreage or size), calls the lee-raleigh connector's pull_owner_mailing_list tool against the statewide NC OneMap parcel mirror and returns a clean CSV of owner names + mailing addresses + site addresses in seconds, deduplicated by mailing address and filtered to private owners (drops government/exempt/HOA/cemetery parcels). Use for any "owners of within of
" / "mailing list for " / "who owns the vacant land near " request. Covers Wake, Durham, New Hanover, Lee, Orange, Johnston, and Chatham counties (NC); coverage grows by county. v1 returns the CSV; Avery labels are a separate skill (gi-plugins #38). --- # Owner Mailing List Produce a deduplicated, private-owner mailing-address CSV from county parcel data. The data work runs **server-side** on the lee-raleigh connector (`pull_owner_mailing_list`), reading a pre-staged statewide parcel mirror — no browser, no extension, results in seconds. ## When to use / not **Use this skill when a broker wants:** - A property-owner mailing list by area and criteria: "owners of 2–5 acre vacant land within 3 miles of 100 Walnut St, Cary NC" - "Mailing list for [area/property type]" - "Who owns the vacant land near [address]?" - Any "owners of `` within `` of `
`" pattern **Do NOT use this skill for:** - **Comps** (internal or external) — use `internal-comps` or `external-comps` - **Single-address owner lookup** — use `owner-lookup` (different skill) - **Demographics** — use `demographics-report` - **Avery 5160 label PDFs** — deferred to gi-plugins #38 (separate skill) - **Phone/email enrichment** — deferred to lee #35/#36 ## Coverage Wake, Durham, New Hanover, Lee, Orange, Johnston, Chatham (NC), served from the NC OneMap statewide parcel mirror. For any other county, tell the broker: > [County name] isn't covered yet. For [county name], go directly to its > county GIS site and export the owner list from there. (The tool answers from whatever counties are staged — when in doubt, run the query; an out-of-footprint address comes back with zero rows or a clear locate error, never a traceback.) --- ## Step 1 — Parse the request Call `helpers.parse_request(text)` on the broker's request string. The function returns: ```python { "subject_property": {"address": "100 Walnut St, Cary NC"}, "radius_mi": 3.0, # None if not specified "size": {"min_acres": 2.0, "max_acres": 5.0}, # {} if not specified "land_class": "vacant", # "" if not specified "raw": "", } ``` **Confirm back to the broker before proceeding:** > Got it — I'll pull owners of [land_class] land, [size range if given], within [radius] miles of [address]. Running the parcel query now. If `radius_mi` is `None`, ask the broker: "How many miles out from [address] should I search?" If `subject_property.address` is blank, ask the broker for the subject address before continuing. Acreage handling: `size.min_acres` / `size.max_acres` may be present, one-sided ("3+ acres" → only `min_acres`), or absent — pass through whichever exist. --- ## Step 2 — Call the MCP tool Call **`pull_owner_mailing_list`** on the lee-raleigh connector: ```json { "address": "100 Walnut St, Cary NC", "radius_mi": 3, "min_acres": 2, "max_acres": 5, "land_class": "vacant", "private_only": true, "dedup_by_mailing": true } ``` - Always pass `private_only: true` and `dedup_by_mailing: true` (the broker contract: private owners, one row per mailing address). - Omit `min_acres` / `max_acres` / `land_class` when the broker didn't give them. - `land_class` accepts: `vacant`, `commercial`, `industrial`, `residential`, `agricultural`. The tool returns `{ok, subject, rows, total_matched, truncated, latencyMs}`; each row carries `owner_raw`, `owner_mail_address`, `address`, `lot_size_acres`, `land_use`, `distance_mi`. An `ERROR:` text response is already broker-legible — relay its substance, never a traceback. --- ## Step 3 — Write the CSV file 1. Map tool rows to CSV rows (deterministic, in the sandbox): ```python rows = helpers.rows_from_mcp(result["rows"]) ``` 2. Write the file: ```python from datetime import date path = helpers.format_csv(rows, request, date.today().isoformat()) ``` `format_csv` derives the flat filename via `default_output_path` — e.g. `owners-100-walnut-st-cary-nc-2026-06-10.csv`. **Report to the broker:** > Done — **[len(rows)] private owners** of [land_class] land within [radius] miles of [address]. > [total_matched] matching parcels; deduplicated by mailing address, government/exempt/HOA/cemetery parcels dropped. > > CSV saved: `[filename]` If `truncated` is true, add: "The list was capped at [len(rows)] rows — tighten the radius or criteria for a complete list." --- ## Output **One CSV file**, short flat filename, written directly to the working directory: ``` owners--.csv ``` Example: `owners-100-walnut-st-cary-nc-2026-06-10.csv` **Rules (load-bearing — Windows 218-char path limit):** - **Never create a subfolder.** No nested paths. - **Never use a long descriptive name.** The slug comes from the subject address only; the filename is generated by `helpers.default_output_path(request, date)` — do not construct it manually. - Brokers run Cowork on Windows where the base output path is already ~125 chars deep; a subfolder + long name tips the total over 218 and Excel refuses to open the file. **CSV columns** (in this order): `owner`, `mail_addr`, `site_addr`, `acreage`, `land_class` Note: `land_class` is the county's land-use code (terse in some counties, e.g. `V` for vacant) — best-effort. `owner` + `mail_addr` (street + city + state + zip) are the load-bearing columns. --- ## Errors — broker-legible only, never a Python traceback | Situation | Message | |---|---| | Tool returns `ERROR: Could not locate ...` | "I couldn't locate [address] — can you confirm the full street address including city and state?" | | Tool returns `ERROR: radius_mi ...` | Ask the broker for a radius up to 25 miles. | | Zero rows returned | "No parcels matched [criteria] within [radius] miles of [address]. Try widening the radius or adjusting the acreage range." | | Rows exist but all filtered | "All matching parcels were government/exempt/HOA/cemetery owned — no private prospects in that area. Try widening the radius." | | Connector unavailable | "The lee-raleigh connector isn't responding — try again in a few minutes." | **Never surface** a Python exception, a stack trace, or a raw tool error payload to the broker.