Check: operate | ComputeSphere Learn

Goal: Confirm you can run a live app with an agent: what it may do alone, when to roll back, and when to take over by hand.

Four questions: three on this module, one on Module 7. Every answer explains itself, right or wrong.

Users report errors right after your agent's latest release. What comes first?Ask the agent for a quick fix and deploy itRoll back to the last good version, then investigateRead the logs until you find the cause

Your agent summarises the logs: 'just a few timeouts, nothing serious'. What do you do before acting on that?Ask for the exact lines, and read the log yourselfTrust it; it read every lineRestart the service to clear the timeouts

Twenty minutes in, the agent has 'fixed' the same failing test three times and now wants to skip it. What now?Let it skip the test so you can shipGive it three more attemptsStop the session and take over

From Module 7: where do you record that your agent must never delete volumes or rotate production secrets on its own?In the instruction file's 'never without asking' sectionIn a message at the start of the sessionIn a skill for deploys

Retry as often as you like.