Operator v0.2.0

Operator v0.2.0

This release makes suggested work easier to organize and adds reusable commands,
live model discovery, and more agent configuration options.

• Organize suggestions by the task that proposed them, edit them in the transcript, and
  accept or start a whole dependency chain at once. Agents ask before filing unsolicited tasks.
• Save reusable prompts as custom slash commands.
• Discover available Claude and Codex models from connected agents, refresh the catalog,
  and pin exact model versions, including Claude Opus 5.5.
• Connect Claude through Amazon Bedrock, including AWS SSO sign-in from the app.
• Choose an opt-in Full access permission mode for Codex tasks.
• Inspect captured agent prompts with the optional prompt-debugging view.
• Improve /clear handoffs and session-versus-lifetime usage reporting; fix missing
  progress indicators after questions and stale conflict-resolution banners.

For Opus 5.5, update your local Claude Code CLI to 2.1.280 or newer. Docker builds
include a compatible version.

Full changelog