Claude Code Remembered Too Much

The real news

The story, straight

A password is a key. A computer command is an instruction.

Put the key inside the instruction, and saving that instruction can save the key too.

Claude Code, Anthropic's coding assistant, asks permission before running most commands. Choose 'Yes, and don't ask again' for an exact command, and it can save that command word for word in .claude/settings.local.json.

Any password inside can remain readable. You asked for fewer interruptions; the permission slip kept a spare key.

Apiiro found secrets in 1,122 of 104,185 public copies of that file on GitHub, where people share code. Those copies contained 1,379 distinct secrets.

The researchers did not test them or establish misuse; 1 notified company confirmed its key had been live. Exposure is established, not a confirmed break-in.

Claude Code's protective ignore rule lives in the user's computer settings, not the project's own exclusion list. It doesn't travel with the project or rescue an already uploaded file.

Replace exposed passwords, tokens and keys first. Then clean current files and saved history, stop tracking the local settings file, and add it to the project's .gitignore.

Keep actual secrets out of commands; use references to securely stored values instead. Deleting a key's photograph doesn't change the lock.

Hardships are coming. I, the Big Dreamer, created this token to help the people.

I truly do see this token going up in value. Ride the rocket with me and let's see where the next 24 months take us.

Buy: https://lilmeme.fun/buy

Repost the warning, not the keys.

Thanks for supporting Lil'meme. Your support is appreciated.