Valt0 v1.2: no app reads your vault until you say yes. Each one gets a prompt, and Touch ID, Windows Hello or your password confirms itβs you. #devtools#security
SOPS leaves the keys readable and encrypts only the values, so a git diff shows which setting changed.
That only holds if you change it with sops edit or sops set. Decrypt and encrypt the file again and every value gets a fresh IV, so the whole file shows as changed.
python-dotenv won't overwrite a variable that's already set. So an old export DATABASE_URL=... in your ~/.zshrc quietly beats the one you just put in .env.
Delete the export and open a new terminal, or call load_dotenv(override=True).
git-crypt has no way to revoke a user. Everyone you ever added still holds the one repo key, and it decrypts new commits too.
File names and commit messages are never encrypted either.
When Claude Code offers "Yes, and don't ask again" for an exact command, it saves that command to .claude/settings.local.json. If the command had a password in it, so does the file.
Apiiro found 1,122 copies of that file with a secret in them on public GitHub.
Your app isn't the only thing reading your .env file. Docker builds without a .dockerignore, backup zips and AI coding agents pick it up too. Worth a look at what's sitting in yours.
Valt0 now works with Java and Kotlin, as well as Node.js, Go and Python.
Your app's secrets stay encrypted and out of config files and logs. Free, no account.
#Java
@0xAndrewShu Deriving the ports from one worktree ID is neat, it keeps all the non-secret config out of the secrets problem. With Parameter Store, Secrets Manager and 1Password all in play, which one wins when a key gets rotated?
@sbintakall Good setup, op run keeps keys out of files. One thing to watch with AI agents: if op is signed in in your terminal, an agent with a shell can run op read as easily as you can. Starting agents from a shell where it isn't signed in closes that.
@Damir_Akaza worth adding: start the agent from a shell with no keys exported. an approved curl can only send what's in reach. that's the part we're working on with Valt0, no plaintext .env in the folder for a CLAUDE.md to ask about
@komal_uk01 the model won't remember it, the chat log will. provider history plus local transcripts (Claude Code keeps them in ~/.claude/projects), so rotate it. not having a plaintext .env for an agent to read is what we're building Valt0 for
@mikalkhoso worth checking the server only serves the build folder. then /.env is a 404 even if the file is sitting right there. the other fix is not having a plaintext .env in the project at all, which is what we're building Valt0 for
@SamVale29 A, as long as .env is in .gitignore before the first commit. the bigger risk these days isn't git, it's everything else that reads the folder: docker builds, zips, AI agents. that's the part we're working on with Valt0, keys stay encrypted next to the code
@ScaleWithMike It's ready and based on your first tweet, you would need the Python client (which we recently released). Hope I understood your needs correctly. If you need help with setting it up, I'm happy to answer any questions or jump on a call.