2.2.4. Generating Scripts and Managing Files

💡 First Principle: Because the CLI is an agent with file and command access, it can move from describing a script to creating, running, and revising it — closing the loop between intent and a working artifact in one place.

A core CLI use case is turning a plain-English goal into scripts and file operations: "write a bash script that backs up these directories and rotates old archives," or "rename these files to a consistent convention and update the references." The CLI can generate the script, create or modify the files, run the result, and iterate on errors — all within the terminal session.

This is genuinely agentic file management, not just text generation. The developer remains the reviewer: you approve actions (especially when Autopilot is off), inspect the diffs, and validate that the script does what you intended before trusting it in a real environment.

✅ Best Practice: For anything that writes or deletes files, review the planned actions before approving, and test generated scripts in a safe location first. The CLI's convenience does not replace the suggest-evaluate-adapt loop.

⚠️ Exam Trap: A generated script that "looks right" can still be wrong or destructive — the same probabilistic-output reality from Phase 1 applies. Validation before execution is the expected responsible behavior.

Reflection Question: The CLI offers to generate and immediately run a cleanup script across your project. What should you check before approving, and which CLI behavior would you keep on to stay in control?