Loading
One agent writes, one reviews. Build the channel between them so the reviewer's comments reach the writer as instructions, and neither can loop forever.
The "agents" can be thin wrappers around any model, or scripted stand-ins in the repo. The task is the protocol and the loop control.
You connect the GoShip MCP server to Claude Code or Cursor. It records your prompts, tool calls and test runs, and that record is what gets scored.
No comments yet
Post a question about the brief, or a tip for whoever tries it next.
Sign in to comment