Describe a blueprint for the scenario and a plan for testing it safely. Return JSON {"scenario_name":"...","steps":["..."],"test_cases":["..."],"rollback":"..."}. steps - numbered steps from trigger to action; test_cases - an array of 3-5 test examples, INCLUDING obviously crooked ones (empty field, different language, double); rollback - how to quickly turn off or roll back the script if it starts to fail. In rollback or in steps describe a silent/test run (the action only logs, not truly sends) before the live launch.
Before setting up, the scenario must be described on paper - this is a drawing that saves hours of repair. The format is simple: numbered steps from trigger to action, and next to it is what comes into the input and what should go out of each step. The testing mechanics are like an airplane: first a test run on the ground. Take 3-5 real examples, including obviously crooked ones (blank field, letter in another language, double application), and run the script on them, looking at the result of each step, and not just the finale. Leverage: test on 'evil' examples specifically - normal ones will pass anyway, everything breaks at the edges. Insider: enable the new script in 'silent mode' - let it work, but the action only logs and does not actually send; We watched it for a day, made sure, turned it on. Second trick: think about the rollback in advance - how to turn off the script with one button if it starts to avalanche. A typical rookie mistake: jump straight into battle on live clients and go off to drink coffee. The first non-standard input and the bot sends out a hundred crooked emails, and clearing up your reputation is more expensive than half an hour of testing.
Unlock access to submit solutions for instant AI review.