Comment on Claude Code deletes developers' production setup, including its database and snapshots — 2.5 years of records were nuked in an instant

• suicidaleggroll@lemmy.world ⁨2⁩ ⁨days⁩ ago

  if you’re denying access to your agentic AI, what is the point of it? It needs access to complete agentic tasks.

  Yes, which it can prompt you for. Three options:

  1. Deny everything

  2. Prompt for approval when it needs to run a command or write a file

  3. Allow everything

  Obviously optional 1 is useless, but there’s nothing wrong with choosing option 2, or even option 3 if you run it in a sandbox where it can’t do any real-world damage.

  source

  • thebestaquaman@lemmy.world ⁨1⁩ ⁨day⁩ ago

    You can fine-grain nr. 2 even more: You can give access to e.g. modify files only in a certain sub-tree, or run only specific commands with only specific options.

    A restrictive yet quite safe approach is to only permit e.g. git add, git commit, and only allow changes to files under the VC. That effectively prevents any irreversible damage, without requiring you to manually approve all the time.

    source
  • artyom@piefed.social ⁨2⁩ ⁨days⁩ ago

    1. Prompt for approval when it needs to run a command or write a file

    And therein lies the problem. You’re giving the LLM control over when to or not to ask for approval.

    source

    • suicidaleggroll@lemmy.world ⁨2⁩ ⁨days⁩ ago

      You clearly have absolutely zero experience here. When you’re prompted for access, it tells you the exact command that’s going to be run. You don’t just give blind approval to “run something”, you’re shown the exact command it’s going to run and you can choose to approve or reject it.

      source

      • wonderingwanderer@sopuli.xyz ⁨1⁩ ⁨day⁩ ago

        Unless you’re managing app permissions on android 🙄

        source