Gotchas, Limits & Anti-Patterns
Gotchas, Limits & Anti-Patterns
The Connector That Lies to You
Connector data can be stale, partial, or filtered without warning.
You ask Cowork for your open invoices. It connects to your accounting tool, comes back with a tidy list of seven, and you start chasing payment.
Two of them were paid yesterday. One large invoice from this morning isn't on the list at all. The number you quoted in a client call was wrong, and you said it with the confidence of live data.
Nothing broke. The connector ran, returned rows, and Cowork summarized them faithfully. The lie wasn't in the answer; it was in the gap between what the connector returned and what you assumed it returned. You read a snapshot as the whole truth.
The fix isn't distrust. It's learning the three ways connector data goes quietly wrong, and asking before you act.
#Stale, partial, filtered
Connector data misleads in three ordinary ways, none of them dramatic.