Secret cleaner

The Secret cleaner finds API keys, tokens, passwords and private keys in a text you are about to paste into an AI chat, a ticket or a message, and replaces each with a placeholder such as <AWS_ACCESS_KEY_1>; the same value always gets the same placeholder, so the text still makes sense. What was found is listed with a masked hint; the values themselves are shown nowhere and kept nowhere.

Fourteen kinds by shape: AWS access and secret keys, Google API keys and service-account private keys, GitHub tokens, Stripe keys, Slack tokens, OpenAI and Anthropic API keys, JWTs, private-key blocks (PEM), URLs with credentials (only the password goes), Bearer tokens, and any value after a word like password, secret, token or api_key.

Anything without a known shape. A password in free text with no password= in front of it, a home-grown token format, a secret split across lines — none of these is caught, and a made-up key of the right shape is. Read the cleaned text before you send it.

A placeholder named for the kind and numbered: <STRIPE_KEY_1>, <STRIPE_KEY_2>. The number keeps two different keys of the same kind apart, and the same value gets the same placeholder every time it appears, so a config or a log still reads right.

Read it, then forget it: it shows the kind, how many times it appeared, and the first four and last two characters — enough to recognise which key it was, never enough to use it. Nothing is stored.

This one takes the secrets out; the AI context cleaner takes the page furniture out. For a config or a log, this is enough. For a whole page, clean it there first, then paste the result here.