Bind each terminal call to its own decoded Evidence ID and require a readable Evidence result before passing the separate nominal workflow grade. A planned URI call with empty content must not pass merely because its name, arguments and status match. Preserve factual and resource grades independently. Validation: 32 targeted context/grading tests, including the reproduced workflow-only false success, and all four original factual/workflow regrades. No unchanged SDK/comparator or model rerun. Co-authored-by: factory-droid[bot] <138933559+factory-droid[bot]@users.noreply.github.com> Co-authored-by: Kaoru0822 <208281322+kaoru0822-kitauji@users.noreply.github.com>