Help centre · Understanding

What does the score on a run mean?

What the score measures, the threshold that holds a merge request back, and what it does not tell you.

The score is the mark the checking agent gives the produced code, between 0 and 1. Below 0.8 the merge request becomes a draft and the pipeline attempts an automatic correction rather than handing you work it considers insufficient.

What exactly is being marked?

How well the produced code matches the issue and the technology's conventions: is what was asked actually done, does it follow the framework's rules, are the modified files consistent with each other.

The score does not measure whether your feature is a good idea, or whether your users will like it. It is a conformance mark, not a relevance one.

What happens below the threshold?

The pipeline runs the developing agent again with the checker's objections, up to the number of attempts your account allows. If the score stays low after those attempts, the merge request is delivered as a draft, with the reasons written as a comment — you see what is wrong instead of discovering incomplete work already merged.

Can I change the threshold?

The 0.8 threshold is set by the vendor and is the same for every account. What you control is the number of automatic correction attempts (Pipeline → Execution): each attempt is a full round trip, and therefore a cost.

Does a low score mean the code is bad?

Not necessarily. A low score means the checker is not convinced. Read its comments on the merge request: sometimes it is right, sometimes it tripped over an internal convention it does not know — in which case the right answer is to write that convention into your contexts so later runs know it.