Skip to main content

Lab 9 - CI: the reference

Lab 9 produces no files in your repository: its result is a pull request on your fork with a triage comment. The transcript records one, with both kinds of comment.

What done looks like​

  • Run tests fails on your push and again on the pull request.
  • Agent triage comments once on the pull request. A second failing push updates that comment instead of adding one.
  • Without a credential, the comment lists the failed tests and names the secrets that would add an analysis. With one, a root cause per test follows the list.
  • The pull request is closed, not merged, and main on your fork is unchanged.

Why it is a good result​

  • The workflow, not the agent, posts the comment, so the agent needs no permission to write to your repository.
  • One comment per pull request, which later failures update, so the pull request stays readable.
  • The summary comes first and works without a key; the analysis is an addition, and its last line names the model that wrote it.

What to debrief​

  • Does the root cause match the break? In the rehearsal it did, for both tests, with the diff's hunks as evidence. But each fix offered two ways out: change the test back, or change the shop to match the test. Only someone who knows why the test changed can choose.
  • The heal suggestion (stretch goal). The rehearsal's run healed the five drifted locators of resources/legacy.resource onto a branch. Review it as in Lab 8: one heal no longer used the keyword's ${GRID} variable, and a second run proposed a different locator for the same line.

Compare yours​

Compare your comment with the two versions in the transcript: the one without a credential, and the one with an analysis.