Spec.cop
Local sample · no inferenceHow it works

Match the ticket
to the diff.

Edit either side and run it again. Every status points to changed text; the score is a visible text-overlap heuristic, not an AI verdict.

What was asked?

Synthetic sample
0 / 12,000

Put each criterion on a line. Prefix criteria with a dash for the clearest comparison.

0 / 12,000

Paste a unified diff. Added and context lines can count as positive evidence; removals are shown separately.

Evidence report

Computed in this browser

What this means: matching words can surface evidence; they cannot establish intent, correctness, or whether a requirement is truly satisfied.

Show your work.

The local score is 70% requirement gap (missing = 1, partial = 0.5) and 30% share of added lines with no ticket terms. Added and context lines can support a match; removed lines are shown separately and never satisfy a criterion. Inspect the evidence before drawing a conclusion.