sk-deep-review

Iterative code review with convergence works by automatically auditing a target project across multiple passes, applying fresh context per iteration and externalizing state, to drive findings toward a stable, resolved state and determine release readiness.

A severity-weighted code audit classifies findings into P0, P1, and P2 priority levels, applies adversarial self-checks on P0 findings, and evaluates three binary quality gates to produce a final release readiness verdict.

You can perform an autonomous code review by invoking the deep-review workflow on your target spec folder or skill package, which automatically initiates the iterative review loop and cross-reference protocol checks.

Yes, automated code auditing tracks quality across multiple iterations by externalizing state and refreshing context per pass, preventing context window decay and enabling recoverable audits until convergence is detected.

The best way to prevent context window decay during long code reviews is to use an autonomous auditing loop that externalizes state and applies fresh context per iteration, ensuring consistent analysis quality throughout the convergence process.

Iterative code reviews for release readiness require a target spec folder or skill package to audit, and convergence depends on successfully resolving severity-weighted findings to pass the three binary quality gates defined in the workflow.