Verify

Acceptance testing for pull requests works by executing predefined checks against your codebase modifications and collecting evidence from test results. An LLM judge then evaluates this evidence against your predefined acceptance criteria to provide an evidence-based decision on whether the change adheres to quality standards.

To automate change validation, provide your user-defined acceptance criteria and optional candidate testing commands. The verification process runs these checks, collects the evidence, and uses an LLM judge to determine if the code modifications meet the specified standards, preventing regressions.

LLM judgment in software quality assurance is an evaluation mechanism where a language model acts as an explicit judge over collected test evidence. It determines whether code changes satisfy predefined acceptance criteria, providing an evidence-based outcome for change validation.

For bounded acceptance verification, you need to provide user-defined acceptance criteria and optional candidate commands for testing. The system uses these inputs to execute checks, gather evidence, and apply an LLM judge to validate that your code changes meet the required standards.

No, code verification with an LLM judge does not strictly require predefined test commands; candidate commands for testing are optional. However, you must provide predefined acceptance criteria so the system has a baseline to evaluate the collected evidence against.