collie-harness:review

You can enforce code review quality gates by running automated Collie rubric reviews on plan documents or code diffs. The review checks 12 red-line violations, answers 10 review questions, and runs an anti-sycophancy self-check to prevent hidden issues and biased approvals.

Automating anti-sycophancy checks is best done by dispatching an isolated Opus subagent to apply the ELEPHANT criteria. This ensures rigorous, objective evaluation of code diffs without sycophantic approvals or developer bias.

Yes, you can use the Collie rubric reviewer during both planning and coding phases. It features automatic mode detection to identify plan, code, or ad-hoc contexts, applying the appropriate quality standards to each phase.

To run a manual review, invoke the review through the Collie auto workflow with your request. The system dispatches an Opus subagent to safely perform the comprehensive checks and returns a fixed-format report with the results.

No, the Collie rubric review does not require any external dependencies or libraries to function. It operates independently by leveraging an Opus subagent to enforce quality gates and return fixed-format reports.

Automated code review misses redundant work because it lacks strict rubric enforcement. Applying the 12 red-line checks and 10 review questions during the planning phase prevents hidden issues and redundant work before coding begins.