WebMCP agent-readiness audit

Assess how well a website works for in-browser AI agents and deliver a scored, prescriptive report. Read-only: this skill changes no application code. When the user wants tools implemented afterwards, hand the report's proposed tool list to the webmcp skill from this plugin, which starts from its own approval gate.

Principles

• Every finding traces to something observed in the repo or the running site; never fabricate gaps or capabilities.
• Prescriptive, not just diagnostic: every gap ships with what to fix first and why it matters.
• Code and analysis stay local; nothing is sent to any external service. The one exception is the optional public scan in step 6: it touches only a public URL the user explicitly approves, never code.

Process