Three work classes achieved a "works" verdict: PR body from a diff (1 of 4), CI log real-or-cascade verdict (0 of 6), and Pydantic model from a spec (3 of 5). Nine classes were marked "works with caveats": release notes from commits (0 of 6), node README (1 of 4), fix a red test (2 of 7), merge conflict (0 of 6), handler bug fix (1 of 4), new ONEX node (0 of 7), RED test from a criterion (0 of 5), and tests for an untested function (1 of 4). Two classes currently "do not work yet": code review (0 of 6) and run a test suite (0 of 5). The primary failures for the caveated classes included rewriting commit prefixes, inventing module paths, breaking constraints, losing lines, misclassifying tasks, naming wrong repos, failing tests for wrong reasons, and using wrong expected values. The non-working classes failed due to speculative false findings and the inability to execute.
