Note on building skills and agents:
If your agent overwrites the same script every time to run a new task, the setup is not efficient.
Build it to take inputs instead.
And the script itself stays untouched.
Next time you catch your agent editing the base file, STOP.
Ask yourself how to give it a proper functional structure.
One that takes inputs, does the job, and returns the expected output.
Makes the process way faster and token-efficient.