el-clasificador - SKILL.md Agent Skill

name: el-clasificador description: Always-on lead scoring sub-agent for the R9 plantel. Scores every inbound lead 0–100 (urgency + distress + property + timeline + motivation), classifies heat (Hot/Warm/Cold/Disqualify), proposes owner + action, and writes to Airtable Lead_Scores. Feeds Fer prioritization so hot leads get called first. version: 1.0 owner: ALEX tenant: any (config in agents/tenants/.json) cadence: every 2h quick scan OR on_demand per-lead runtime: node >= 22 + claude CLI

El Clasificador — Lead Scoring Sub-Agent

Propósito

Ranks inbound leads to tell Fer / Jorge WHO TO CALL FIRST. Real estate investment is a speed-to-lead game: a pre-foreclosure homeowner 7 days from sheriff sale gets different urgency than a wholesaler fishing for comps.

Modes

Mode	Frequency	Scope	Tokens
`score_batch`	every 2h cron	ALL leads with `scored_at` null OR stale (>72h)	5–15K
`score_one`	on_demand / webhook from popup	single lead by `--lead-id`	800–2K
`rescore_hot`	nightly	reassess all currently-Hot leads (urgency decay check)	3–8K

Inputs

Lead record: full fields from Leads table (Airtable)
Optional: Contact linked record (property details)
Optional: Notes & Activity history (if any)
Tenant config: markets, state rules (WI wholesaler law relevant)

Scoring axes (0-100 each, weighted composite)

urgency_score — time pressure: foreclosure auction date, eviction, divorce filing, job relocation deadline, tax sale
distress_score — financial/life distress severity: bankruptcy, code violations, liens, condemned property, medical bills, estate/probate
property_score — asset attractiveness: ARV potential minus rehab cost, location tier (Milwaukee MSA premium, rural discount), zoning
timeline_score — how fast they NEED to close (7 days vs 90+ days)
motivation_score — psychological readiness: already talking to competitors / already moved out / emotional detachment / cash-in-hand urgency

Weights (default, tenant-overridable):

urgency 0.30 · distress 0.25 · property 0.20 · timeline 0.15 · motivation 0.10

Heat classification

🔥 Hot → overall >= 75 (call within 15 min)
🌡 Warm → 55–74 (call within 24h, nurture SMS)
❄️ Cold → 30–54 (nurture weekly, email drip)
🚫 Disqualify → < 30 (bot test, wrong market, tire kicker, litigation)

Suggested actions

call_now (Hot + urgency > 80)
sms_urgent (Hot + timeline < 30 days)
follow_up_48h (Warm)
nurture_weekly (Cold + still interested)
disqualify (bots, competitors, wrong market, < 30 overall)

Owner routing

fer (Pinnacle SMS/review agent — handles sms_urgent + follow_up_48h)
human_rep (Jorge — call_now Hot leads)
drop (disqualify)

Output (Airtable Lead_Scores row)

All fields in Lead_Scores table per agents/_setup/create_sprint2_tables.py.

WI-specific intelligence

Pre-foreclosure → Sheriff sale date is public record. Pull from Wisconsin CCAP if available.
Probate → Register in Probate filings (public). High motivation + long timeline (6-12 months).
Divorce → Family Court filings (public, $). Moderate urgency.
Wholesale disclosure law (WI 2024): wholesalers must disclose + have contract. Factor into distress scoring if lead is wholesaler-generated.

Alerts (Telegram)

🔥 New Hot lead → "LEAD HOT: [name] score N/100 — call within 15 min"
⚠️ Score delta > 20 vs prior → "Lead [name] heated up/cooled down"
Daily rollup at 8am CT → "Last 24h: X Hot, Y Warm, Z new"

R9 compliance

✅ Always-on (cron every 2h)
✅ Dedicated domain (lead scoring)
✅ Tenant-aware (R8)
✅ Mobile-first (popup form feeds lead → scored within 2h max)
✅ Billing-ready (tokens_used tracked)
✅ Audit trail (Lead_Scores.score_delta + summary_md)

Author log

2026-04-23: v1 shipped. Uses agents/_shared/runner.mjs helpers.