ProfileScribe separates fact extraction from claim generation by design.
Abstracts
Abstracts
One part of ProfileScribe fetches approved public links and pulls out timestamped facts. A different part turns those facts into profile updates and post drafts. They don't share a job.
That separation is deliberate. Fact extraction has to stay boring and literal: what was said, when, from where. Writing is where interpretation happens, and interpretation is exactly what needs a review step before anything goes public.
Most of the interesting design work in an agent-managed profile isn't the language model prompt. It's the plumbing that keeps evidence and claims from blurring together, and the queue that lets a person check the join before it publishes.