feat: restructure header to match ArtifactFS style

Update header with bold h1, CI badge, blockquote notice, and clean paragraphs.

Co-authored-by: Sree Narayanan <23213899+eersnington@users.noreply.github.com>

Author: v0

README.md

@@ -1,17 +1,12 @@
-<h3 align="center">agent-container</h3>
-
-<p align="center">
-  Give your agents tiny boxes, powered by <a href="https://github.com/cloudflare/workerd">workerd</a><br />
-  Agent Container is a workerd-based sandbox designed to give coding agents isolated execution environments, providing structured bindings (workspace, exec, env) scoped to a single repository rather than the host system. It's ideal for AI coding tools, agent frameworks, and platforms that need to run untrusted code with fine-grained capability control.
-</p>
-
-<p align="center">
-  <a href="#quick-start">Quick Start</a> &middot;
-  <a href="#why">Why</a> &middot;
-  <a href="#architecture">Architecture</a> &middot;
-  <a href="#api">API</a> &middot;
-  <a href="#development">Development</a>
-</p>
+# Agent Container
+
+[![Build & Test](https://github.com/vercel/agent-container/actions/workflows/ci.yml/badge.svg)](https://github.com/vercel/agent-container/actions/workflows/ci.yml)
+
+> This project is under active development. APIs may change.
+
+Agent Container is a [workerd](https://github.com/cloudflare/workerd)-based sandbox for running untrusted code from AI agents. It exposes structured capability bindings (filesystem, exec, env) scoped to a single repository, rather than giving agents raw access to the host system.
+
+It's useful for coding assistants, autonomous agents, and any platform that needs to run agent-generated code with fine-grained control over what that code can access.
 
 ---