vc-autoresearch

A self-contained autoresearch loop automates gap discovery, report writing, issue fixing, and outcome rechecking until metrics improve or a cap is reached. It enforces structured iteration reporting and deterministic configurations for repeatable improvements.

An autoresearch loop is a structured workflow orchestration primitive that finds gaps, writes reports, applies fixes, and repeats until quality metrics improve. It serves as shared bookkeeping for multi-phase plan-validate-fix and execute-validate-fix cycles.

Yes, the loop explicitly drives plan-validate-fix (PVL) and execute-validate-fix (EVL) cycles. It provides shared bookkeeping and unified orchestration across these phases for consistent quality improvements.

You ensure repeatable improvements by enforcing structured iteration reporting, guard checks, and deterministic configurations during the fix and recheck phases. This prevents regression and guarantees quality metrics trend upward until targets are met.

The best way is using a shared loop primitive that coordinates gap discovery, fix application, and guard checks across multiple phases. This approach maintains consistent bookkeeping between validation cycles to ensure reliable outcomes.

You should cap the loop when quality metrics reach the desired target or when the maximum iteration limit is hit. Guard checks enforce these boundaries to prevent infinite cycles and ensure the workflow terminates deterministically.