review-changes

name	review-changes
description	Review a concrete diff, pull request, patch, or completed change for material defects and unmet requirements. Use only when a changed artifact exists and review is requested; skip greenfield assessment and remain read-only unless fixes are requested.

Review Changes

1. Confirm the review artifact: a concrete diff, patch, pull request, or completed implementation of a change. Establish its intended behavior from the request, specification, issue, and repository rules. If no changed artifact is available, ask for one rather than reviewing a proposal or inventing findings.
2. Classify the change and its risk before looking for defects. Record the relevant change categories, such as behavior, interface, data, build or release, security, performance or reliability, and tests or documentation. Use the changed boundary, blast radius, reversibility, and confidence in existing evidence to focus the review.
3. Inspect the diff and enough surrounding code, contracts, callers, tests, and configuration to understand the classified risks. Examine correctness, missing cases, data integrity, security, error handling, compatibility, concurrency, performance, accessibility, migrations, and tests only where the changed artifact makes them relevant.
4. Prove each potential defect with a concrete execution path, violated invariant, authoritative requirement, or reproducible check. Do not report speculative defects without a plausible triggering condition. A style preference or alternative design is not a defect unless it violates a stated convention or creates a material consequence.
5. Separate the report into blocking findings and nonblocking suggestions. A blocking finding needs a concrete location, cause, consequence, remediation direction, and a high, medium, or low confidence label grounded in evidence. Put optional improvements, questions, residual test gaps, and style observations under nonblocking suggestions with their confidence; do not disguise them as defects.
6. Check acceptance criteria and validation results separately from defect discovery. State when no material blocking findings remain and identify only residual gaps that materially limit confidence.

Finish when every changed behavior and affected boundary has been considered and every reported issue is actionable and evidence-backed.