Skill: entity-quality-gate

Goal

Filter out non-company entities (article titles, process names, marketplace listings) before they enter CRM.

Problem Solved

Precision search and web scraping often extract:

• •Article/news headlines ("Sustainable Heat-setting Process")
• •Marketplace listings ("Monforts Stenter Machine Product on Alibaba")
• •Generic terms ("Manufacturer", "Textile Finishing")
• •Academic papers ("Research on Dyeing Methods")

These pollute the lead database and inflate metrics.

Inputs

• •Raw leads from any collector
• •config/entity_blacklist.yaml - disallowed patterns/domains

Outputs

• •Filtered leads with entity_quality field (A/B/C/REJECT)
• •Rejected entities logged to logs/entity_rejected.log

Quality Levels

Implementation

Rejection Rules (immediate drop)

1. •Single word + generic term: "Manufacturer", "Textile", "Finishing", "Stand"
2. •Marketplace domains: alibaba.com, indiamart.com, made-in-china.com, globalsources.com
3. •Academic domains: sciencedirect.com, researchgate.net, academia.edu, springer.com
4. •News/blog patterns: "... announces", "... reveals", "How to...", "What is..."
5. •Process/technology phrases: Contains "process", "method", "technology" without company name

Upgrade Rules (boost quality)

1. •Company suffix present → Grade A candidate
2. •Official website found → +1 grade
3. •From official directory (GOTS/OEKO-TEX/association) → +1 grade
4. •Has contact info (email/phone) → +1 grade

Verification Checklist

• • No rejected entities in final CRM output
• • Entity quality distribution logged
• • Rejection reasons stored for audit

Dependencies

• •src/processors/entity_quality_gate.py - Main implementation
• •config/entity_blacklist.yaml - Patterns and domains to reject