Using Virillio Code
Terminal and tools
Understand how the agent moves from a conversation to actions in your project.
On this page
How tool activity works
The agent can use available tools to read files, search the project, edit code, and run commands. Tool activity appears as part of the work so you can inspect what happened.
A tool result is evidence about an action, not necessarily proof that the whole feature works. For example, a successful build does not prove that a form behaves correctly in the browser.
Run a focused command
Ask for the outcome and the project context:
Find the test command documented by this project and run the tests
for the search feature. Report failures without changing dependencies.Review the working directory, command, and requested permissions. If a command needs a runtime, package manager, or service that is not installed, resolve that prerequisite before retrying.
Use the terminal alongside the agent
Use the workspace’s terminal controls when you want to inspect output or run a command yourself. Avoid editing the same file or launching the same long-running process simultaneously without telling the agent.
Tell the agent about manual changes before it continues. Its earlier view of the project may no longer match the files on disk.
Handle failures
Read the first meaningful error. Check the directory, missing dependencies, and required local services. Share a redacted error message with the agent rather than repeatedly approving the same failing command.
Commands can affect systems outside the project. Treat deployment, deletion, database migrations, and publishing as actions requiring a deliberate decision.