trident

Evidence-based code review scans changes with multiple lenses, then independently verifies findings before rendering a final judgment. This three-stage pipeline of scanning, verifying, and arbitrating surfaces defects with high confidence and reduces false positives by requiring proof for each flagged issue.

You can automate PR review by running the skill across multiple scopes including unstaged, staged, all-local diffs, commit ranges, and directory audits. It supports quick triage for small diffs and deep reviews for complex changes, automatically cleaning up worktrees after the operation completes.

Yes, you can run automated security and code quality checks on local diffs safely. The operation is non-destructive, using confirmations and stop cues to prevent unintended changes, and includes automatic worktree cleanup to keep your repository intact during multi-lens scanning.

The best way to review large codebases is using a deep review mode that scans for architectural smells like SOLID violations and dead code alongside security and data integrity. The multi-lens scanner surfaces these quality signals, which an independent verifier confirms before a final verdict is delivered.

Yes, multi-lens code review supports directory audits and commit range analysis alongside unstaged, staged, all-local, and PR diffs. It structures the output as an auditable YAML contract, allowing you to trace findings from the initial scan through independent verification to the final judgment.