Step 1 – Define the Task

Imagine you want a script that archives old log files.

Start with:

Create a Bash script for an authorized Linux server that:

- Finds .log files older than 30 days
- Archives them into /var/archive
- Does not modify files outside the intended directory
- Handles filenames containing spaces
- Reports errors

Step 2 – Add Security Requirements

Now make the requirements explicit:

Make the script security-conscious.

Include:
- strict error handling
- safe quoting
- input/path validation
- checks that required directories exist
- clear error messages
- no destructive operations without validation

Explain the security reason for each measure.

This usually produces a much better result than simply asking:

“Write me a Bash script.”

Step 3 – Ask AI to Review Its Own Script

Paste the generated script back to AI:

Perform a security review of this Bash script.

Look for:
- command injection risks
- unsafe variable expansion
- path traversal
- dangerous rm/mv/cp usage
- race conditions
- permission problems

For every issue, explain the problem and show a safer alternative.

Step 4 – Test It Safely

Before running a script against real files, test it with a temporary directory:

mkdir -p ~/bash-lab/logs
touch ~/bash-lab/logs/test.log

Then run your script against the test directory.

Check what it does before pointing it at production data.

The Workflow

Task
 ↓
AI Generates Script
 ↓
Security Requirements
 ↓
AI Security Review
 ↓
Lab Testing
 ↓
Production

AI is great at generating shell code, but Bash gives you enough power to cause real damage with a small mistake.

Always review and test generated scripts before running them on important systems.