johnnycrypto289/masonclawv4/prospect-scraper

SKILL.md

prospect-scraper

Purpose: produce usable contractor leads daily with a strict fallback chain and scored outputs.

Required Output

CSV file at memory/prospects/YYYY-MM-DD.csv
Summary file at memory/prospects/YYYY-MM-DD-summary.md

CSV columns:

date
business_name
trade
city
state
phone
website
source
pain_signal
score

Minimum Quality Bar

At least 20 leads per run
Each lead has business name + at least one contact path (phone or website)
Score each lead 1-10
Include top 10 in summary

Fallback Chain (Must Execute in Order)

search-based discovery
direct fetch/parsing
browser navigation capture
curl with browser-like headers
script-based extraction (requests/bs4)
source rotation (Google Maps, Yelp, BBB, Angi, Houzz, Thumbtack, YellowPages, Manta)
query rotation (city + trade + intent phrase)

If blocked, log every attempt and why it failed before reporting blocked.

Scoring Rubric (1-10)

+3 explicit missed-call/slow-response signal
+2 evidence of ad spend / growth push
+2 weak web stack / no online booking
+2 size-fit (small-mid contractor likely owner-led)
+1 local service area fit with current target geography

Operating Rules

Do not stop at first source failure.
Keep trying until quality bar is met or all fallback levels fail.
Save raw notes to memory/prospects/YYYY-MM-DD-raw.md.