Six Claude Code plugin bundles (writing, research vault, explainer videos, NetLogo, dev tools, life) and seven multi-agent workflows, one-command install.
LLM customer with a hidden goal, 13-tool agent, reward gated on the database diff before a judge checks consent and honesty. Red-teaming caught a forged-consent hole in the judge.
Hackathon agent for incident response: a GPT-5.5 supervisor calls code, company, memory and action subagents, then opens a real PR, Gmail draft and Slack post.