I instructed agents observing other agents at runtime and it resulted to several documentation fixes, that multiple `/review` runs didn't catch. Essentially, t

I instructed agents observing other agents at runtime and it resulted to several documentation fixes, that multiple `/review` runs didn't catch. Essentially, the prompt was `apply TDD or Nielsen's user testing to agents-observing-agents via tmux or subprocesses`. Details are on the blog post and GitHub repo.