Show a dry run and ask before destructive commands

How to do it

Agents and tired people run commands they meant to check first. A dry run shows the exact list of records, files or messages, with a count. Make the dry run use the same code path as the action, so the list matches what happens. Without a terminal, the confirmation follows never wait for input when no person is there. Infrastructure tools follow the same idea with a plan, as review the plan and apply only from CI says.

Examples

Python example

def delete_drafts(older_than_days: int, dry_run: bool, assume_yes: bool) -> None:
    drafts = invoices.drafts_older_than(older_than_days)
    for draft in drafts:
        print(f"would delete {draft.number}" if dry_run else f"deleting {draft.number}", file=sys.stderr)
    if dry_run or not confirm(f"Delete {len(drafts)} drafts?", assume_yes):
        return
    invoices.delete_many([d.id for d in drafts])

Go example

func deleteDrafts(olderThanDays int, dryRun, assumeYes bool) error {
	drafts := invoices.DraftsOlderThan(olderThanDays)
	for _, d := range drafts {
		fmt.Fprintf(os.Stderr, "would delete %s\n", d.Number)
	}
	if dryRun || !confirm(fmt.Sprintf("Delete %d drafts?", len(drafts)), assumeYes) {
		return nil
	}
	return invoices.DeleteMany(ids(drafts))
}

How to check

A test runs each destructive command with --dry-run on a test copy, and nothing changes. The list it printed matches what the same command then changes with --yes.

Related rules

• DSK-23 Never wait for input when no person is there in Desktop apps, extensions and CLIs
• INF-3 Review the plan and apply only from CI in Infrastructure and environments