← LibraryBuilder · B9
Reviewer & QA agents
Build-along · ~12 min at your own keyboard
An agent that just wrote a brief is soaked in its own reasoning, the same bias as humans reviewing their own work. A reviewer subagent with a clean context reads the output cold and catches what the writer literally cannot see. For a brief, point it straight at the record: proposed is not agreed, and every fact cites the message it came from.
Do this
0/4Share what happened
Paste the reviewer's findings on a real brief and which ones were legit.