The right environment
Commands run on the selected device with its installed tools.
Let AI run tests, builds and diagnostics on the right machine. Enable terminal access separately on each device.
On Linux dev, in /srv/project, show the test command first. After approval, run it and check the exit status.
Commands run on the selected device with its installed tools.
Specify the project directory so commands target the correct repository.
Start a process and read its output until completion.
Enable Allow programs and commands on the target device and review its OS account permissions.
Ask the AI to show the command and working directory before system changes or deployments.
After execution, inspect output, exit status and test results. Long sessions need subsequent output reads.
On Linux dev, in /srv/project, show the test command first. After approval, run it and check the exit status.
File permissions do not constrain the shell. Commands have the OS user capabilities. Use a separate user, container or VM for tighter isolation.
Run short checks synchronously. Start a build or development server as a managed process and inspect its output.
Do not paste passwords or tokens into ordinary AI requests. Use the project’s existing secure environment configuration.
No. Enable it locally and independently of folder access.
Not through shell tools. Without terminal access, AI can read code and suggest commands for you to run.
Use process-session controls. Terminating unrelated OS processes needs a separate local permission.
Choose a working folder and AI client. Add writes and terminal access as your task requires.