WebMCP: add explicit human-to-agent handoff

[erinepshovel-code]

Purpose

Complete the human/agent gestalt loop on /webmcp/ without pretending the current WebMCP spec has a page-to-agent prompt API.

Changes

• add one ordinary-language textarea: what the human wants changed or produced
• no internal skill identifiers are typed by the human
• nothing is sent merely by typing
• after explicit Send, dynamically register read-only tiw_human_handoff
• handoff returns the exact selected skill, dependency-first required skill set, registry provenance, human request, and authority/persistence boundaries
• changing the selected skill or editing the request invalidates the published handoff until Send is pressed again
• human text remains page-session-only; it is not stored or posted to the public remote MCP server
• the five registry tools remain the base surface; the sixth handoff tool exists only after explicit human Send

Boundary

WebMCP makes the sent handoff agent-visible, but the web standard does not allow the page to force the browser agent to invoke it or inject a prompt into the agent.

[chatgpt-codex-connector]

Codex Review Summary

This comment shows the latest Codex review activity on this pull request.

Review	Status	Commit	Review trigger
📝 Code Review	✅ Completed 2026-09-02T11:21:06.444192Z	db10d02	PR opened

ℹ️ About Codex in GitHub

Your team has set up Codex to review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review" or "@codex security review".

Codex reacts with 👀 while any review is running, comments if it has suggestions, and reacts with 👍 once all reviews finish with no findings.