.png&w=256&q=75)
🤓 Latest Submissions
Verity
A passing test suite does not always mean a developer task is complete. Tests may miss a boundary case, or a requirement may never have been tested at all. Reviewing every acceptance criterion against the implementation takes time and can leave gaps before release. Verity gives developers a repeatable way to check that work. They write a task specification with labelled acceptance criteria, then use IBM Bob to draft executable checks against their code. Verity runs those checks, compares their results with the criteria in the specification, and generates JSON and Markdown reports showing what is VERIFIED, FAILED, or UNCERTAIN. If a criterion has no check, Verity marks it UNCERTAIN rather than reporting that the task is complete. The repository includes a copyable TypeScript template and examples covering a CSV export, pagination, and string utilities. In the CSV example, all 73 ordinary tests passed while Verity’s separate audit probe found a missing requirement. After the audit behavior was repaired, the same probe passed and the task reached 8/8 verified. Developers can follow the template to add checks for their own tasks and rerun them after a fix. Verity is a prototype: Bob helps create the task-specific checks, and Verity executes and reports them. Developers review the checks before trusting the result.
27 Sep 2026
