Scripts you can trust.
WHAT A RELIABLE SCRIPT DOES
States what it is doing Stops on the first failure Leaves things consistent if interrupted Is safe to run again Logs enough to diagnose
WHAT TO SET AT THE START
Strict failure behaviour, so an error stops the script rather than continuing.
WHY
Otherwise a failed step is followed by steps assuming it succeeded.
WHAT TO CHECK BEFORE STARTING
Preconditions: required files, permissions, connectivity, disk space
WHAT TO BUILD IN
A mode that reports without acting Confirmation for anything destructive Bounds on how much a single run may affect
WHAT TO AVOID
Assuming a command succeeded Assuming a path exists Hard-coded paths and hostnames
WHAT TO PARAMETERISE
The environment, so the same script serves each.
WHAT TO RETURN
A meaningful exit status.
WHY
So whatever called it can tell whether it worked.
WHAT TO TEST
The failure paths, by making things fail deliberately.
