Can the Snaper agent run arbitrary commands on my machine?

No. The agent does not expose a shell, an SSH session, or any generic “run command” endpoint. It only reads instructions from your private Datashelter S3 bucket and executes a fixed, hard-coded set of actions.

What the agent can do

When the agent polls its bucket, it will only act on these command types:

• backup — trigger a configured file or database backup
• list — list existing backups
• delete — delete a backup from the index
• browse — build or refresh the remote browse index so you can explore your backups from the dashboard
• db_discover — probe local database engines to list reachable instances
• configure_db — refresh the database connection parameters stored in your config
• update — upgrade the Snaper binary to the latest released version

Any other payload is ignored.

Why this is safe

• The agent never opens an inbound port. It only makes outbound calls to your S3 bucket and to Datashelter’s API.
• The command list is hard-coded in the binary. A malicious payload cannot ask the agent to run rm -rf or any other shell command.
• Every command is signed and scoped to your server. The agent refuses anything that does not match its own identity.
• You can audit exactly what the agent is doing at any time with snaper agent status and the logs under /tmp/snaper.