review-fix

Automating code-review fixes is done by loading a review artifact, filtering findings by severity, and batching fixes in a dependency-safe order. The process dispatches fixer agents in parallel, validates changes, and updates the artifact in place with Fixed or Failed statuses.

Yes, dry-run options are supported for code-review fixes. A dry run plans the dependency-safe batches and simulates the fixer agent dispatch without applying changes, allowing you to review the proposed task graph and per-findings fix log before execution.

Worktree isolation for code-review fixes ensures batched changes are applied in a separate, clean environment. When enabled, the system isolates the fixer agents and validates changes within the worktree boundaries before writing the final fix report and updating the review artifact.

Filtering code-review findings by severity is handled by parsing the review artifact and grouping results into deterministic, parallelizable batches. This ensures fixer agents can execute safely in parallel without file or dependency conflicts during the fix process.

You need a review artifact path or PR number to automate code-review fixes. The system loads and parses the artifact from local worktrees or PR branches to extract findings, filter them by severity, and orchestrate the automated fix batches.

Batched code-review fixes may fail if dependency-safe ordering is violated or validation detects issues after applying changes. The system logs these failures per-findings in the generated fix report and updates the review artifact in place with Failed statuses for auditing.