Skip to main content
Plain coverage is structural: a requirement with at least one linked test is “covered”, even if a single trivial smoke test stands in for five acceptance criteria — false confidence. Turn on “Assess coverage depth” (opt-in, off by default — when off, the Coverage tab is strictly unchanged). After Analyze coverage, a Depth column appears with a per-requirement band: Expand a row to see the findings (the reasons, with links) and concrete improvement axes — “add a negative test for…”, “flesh out the tests without steps”, “confirm a test covers X”.

​
Deterministic & advisory

The analysis is 100% deterministic and AI-free — it counts acceptance criteria, measures step depth, looks for negative / edge keywords (FR + EN), and checks term overlap and execution. It is advisory only: it never blocks anything, never feeds the quality score or GO/NO-GO, and never changes an outcome. The wording stays cautious (“possible”, “apparent”) — it is a review this signal; you decide. It is read-only (no new API call or scope) and degrades gracefully per requirement: a test with no steps still counts, and no acceptance criteria yields Underspecified — never an invented verdict.

Quality Gate — grade & GO/NO-GO

The deterministic score.

The Test coverage tab

Depth on the work item.