project-review

Automating architectural conformance checks involves running deterministic mechanical checks across the repository followed by an LLM judgment pass to identify design drift. This ensures code adheres to structural invariants before heavy-weight code reviews.

An LLM judgment pass in code review uses language models to evaluate non-deterministic judgment invariants. It complements mechanical repository audits by catching subtle architectural conformance issues and design violations that static rules cannot detect.

Yes, running repository design verification requires both Python and Claude Code context. These dependencies are necessary to execute the scripts that perform mechanical checks and the LLM judgment pass for architectural conformance.

The best way to prevent design drift is implementing a conformance gate that combines deterministic mechanical checks with LLM judgment. This verifies code against design invariants quickly, catching gross violations before manual code review.

Yes, repository audits can evaluate non-deterministic design rules using the LLM judgment pass. While mechanical checks handle deterministic repository constraints, the LLM judgment assesses complex architectural conformance and design invariants.