Detect and consume WebMCP declarations from sites

[lonexreb]

Summary

Auto-detect sites that publish W3C WebMCP declarations (navigator.modelContext) and consume them directly — skipping traffic capture entirely.

Context

From RESEARCH-DEEP-DIVE.md: WebMCP is an emerging standard where websites declare their AI-accessible capabilities. If a site publishes WebMCP, site2cli can import the declaration instead of inferring capabilities from traffic.

Implementation

• During discovery, check for navigator.modelContext via JavaScript eval
• Parse WebMCP declaration into site2cli's internal model
• Register as Tier 3 (direct API) immediately — no promotion needed
• Fall back to traffic capture if no WebMCP found

Files

• src/site2cli/discovery/capture.py — WebMCP detection during page load
• src/site2cli/registry.py — register WebMCP-sourced sites at Tier 3