SCOPE — Structural Coverage & Observability Platform Engine
Run only the tests your change can break.
Full regressions burn hours of CI on tests that cannot fail. A change touches only a sliver of the suite — SCOPE finds that sliver, proves it, and reports what nothing covers.
How it works
Snapshot once
Code plus per-test coverage from a green run — which test executes which line.
Map the change
Your edit is resolved down to the exact impacted lines.
Get the delta list
Every test covering an impacted line runs. The rest are skipped.
Name the gaps
Every impacted line that no test reaches is reported — never silently skipped.
No silent blind spots
Test selection alone is not enough. If a change lands on code that nothing exercises, a smaller test run will still come back green — and the risk ships.
Every SCOPE run therefore reports coverage as well as selection: how much of the change is covered at all, and the specific impacted lines that no test reaches. A gap is always named. It is never quietly dropped from the list.
That named gap is exactly what GNOME consumes next.
Ask your assistant. Ship the answer.
SCOPE ships vst, an MCP server that puts the whole pipeline behind a conversation — in Claude Code, or any MCP-capable client. You ask which tests a change needs; the assistant drives the real tools and returns the tools’ own results, not a guess.
Ask it to run them, and it runs them. Ask it to double-check against the full suite, and it proves the selected list was a superset of what actually mattered.