P2: web crawl markdown needs chrome strip + page-specific titles before index

[abojja9]

Summary

Web crawl→markdown pilot (8/8 Moss docs pages) produced usable text but repeated documentation chrome in every page and site-generic titles, not page-specific. Bad for indexing quality without normalize step.

Evidence

• moss-web-crawl-eval customer scorecard (Firstmate data)
• Cookbook path; product path is Ash Add benchmark suite comparing Moss vs Pinecone, Qdrant, and ChromaDB #54 (NO-GO separately)

Expected

Boilerplate stripping + canonical title extraction before index; metrics for chrome ratio.

Priority

P2 product quality for web KB (P1 if selling web ingest now).