vishalgojha skills on Remote OpenClaw

action-suggester

Generates ranked follow-up action suggestions (call, email, or visit) from lead summaries or lead lists. Designed to sit in a supervised pipeline where a human or supervisor approves suggestions before any action is taken. Does not execute actions, send messages, or write to any system.

lead-storage

Persists validated lead objects to Google Sheets or a database after a supervisor confirmation token is verified. Enforces idempotency by lead_id to prevent duplicate inserts from repeated broker forwards. Preserves metadata fields for location, priority, and extraction context alongside each stored record.

lead-extractor

Extracts structured lead records from parsed WhatsApp real-estate messages. Classifies each message as an inventory listing or buyer requirement, and pulls name, phone, budget, and property details into a validated schema. Designed to run after message-parser and before india-location-normalizer.

sentiment-priority-scorer

Scores normalized real-estate leads across sentiment, urgency, intent, and recency dimensions to produce a deterministic 0-100 priority score and P1/P2/P3 bucket assignment. Operates read-only — no writes, messages, or outbound actions. Designed to sit in a multi-agent chain after lead normalization.

message-parser

Parses raw WhatsApp chat exports in TXT or JSON format into a normalized array of message objects, each with a timestamp, sender, and content field. Handles multiline continuation lines, timestamp collisions, and system boilerplate like group creation notices or encryption messages.

india-location-normalizer

Normalizes raw Indian real-estate location text into structured city, locality, and micro-market fields for Mumbai and Pune leads. Uses a curated alias lookup map to resolve shorthands like Andheri W, PCMC, Hinjewadi, and Wakad. Sets confidence scores and flags unresolved matches rather than guessing.