Drawgent wires your Claude Code or Codex to a live Excalidraw canvas
An open Show HN turns a whiteboard into an agent workspace — plus Claude Code's new audit and gateway controls, and a chess-analysis skill worth forking.
Copy markdown
Your coding agent, now living on a whiteboard
Drawgent connects your own Claude Code, Codex, or opencode to a live Excalidraw board over an agent-protocol (ACP) bridge. Ask in a chat panel or drop an `AGENT:` note on the canvas, and it reads a screenshot plus the scene data, edits, and verifies its own work. It runs locally at 127.0.0.1:7300 against your repo — a weekend build for anyone who sketches architecture before they write it.
Claude Code 2.1.283 adds audit and gateway controls
The latest Claude Code ships gateway, MCP, plugin, workflow, and /doctor audit controls, plus a new prompt-audit capability and managed model / blocked-model settings — alongside faster startup and quicker session resume. The startup and latency wins land for everyone; the audit surface matters most if you run agents behind a gateway.
A Claude Code skill that reviews your chess games
brumar/chess-postmortem-skills turns Claude Code into a chess coach: feed it a Lichess PGN and it runs Stockfish on every move, then explains the blunders in plain English — it can even render a narrated MP4 with an eval gauge. MIT-licensed, and a clean template for packaging any 'analyze-then-explain' workflow as a shareable skill.