1 min lesson
Always-on agent
Complete "Always-on agent". Check the final state against the lesson before you move on.
Check your understanding
Run this practice: /automate — draft one automation from a proven run.
Outcome: The first run opens a pull request, so a human reads the change before it lands.
/automate — draft one automation from a proven run
Always-on agent
SayA reviewer asked for a change on my PR. I made the fix by hand, the checks passed, and I kept the session. Turn it into a repeatable automation.
DoRun /automate on the successful local agent session.
Type
Trigger: GitHub PR review commentDoPick the PR review comment trigger from the GitHub trigger list.
SeeThe automation now fires when a reviewer leaves an inline comment on a pull request diff.
Type
Prompt: When a review comment requests a change, reproduce it, apply the smallest safe fix, and run the checks before proposing it.SeeThe drafted prompt is more specific than the original chat.
DoKeep the default output (open a pull request) and confirm the tool and PR permissions match the workflow.
SeeThe first run opens a pull request, so a human reads the change before it lands.
Learn more
Optional practice
Practice: Always-on agent
QYou've applied it. What actually proves the work is done, not just a plausible answer?