The secret scanner I run before anything goes into AI memory
I almost loaded 8 passwords and tokens into a memory all my bots can read. Now every note gets scanned first.
What it looks for
Known key formats: things that start like sk-, xai-, AKIA, ghp_, AIza, or long three-part tokens.
Labelled secrets: anything after "password:", "token=", "api key:" or "secret:".
Random-looking strings on lines about logins: long mixes of upper case, lower case and numbers near words like "login" or "credential".
What it does
Replaces each one with [REDACTED] before the note is saved, and counts how many it found.
The habit
Scan before anything goes into AI memory, every time.
Change any password or token you find, even an old one.
Don't rely on the scanner alone. Keep secrets out of notes in the first place.