plan-eval

To evaluate an architecture plan, ingest the plan document to cross-check consistency with existing codebases. This process returns structured findings highlighting severity, location, and recommended actions for gaps in soundness and executability.

Parallel agent evaluation is a method that applies multiple expert lenses, such as best practices and risk management, simultaneously to a plan. It orchestrates reviews to surface critical gaps and returns prioritized issues with concrete recommendations.

Yes, multi-agent plan review supports large architecture proposals by coordinating parallel evaluation lenses. It scales with plan size, ingesting the proposal to identify completeness issues and cross-file linking errors across the project.

Checking plan-to-code alignment involves ingesting the plan and cross-referencing it against existing codebase conventions. The evaluation returns structured findings detailing any misaligned implementation, including severity and location of inconsistencies.

The best way to structure risk management findings is using a standardized format that includes severity, location, rationale, and recommended actions. This ensures prioritized issues from the evaluation are actionable for the development team.

Yes, you need to provide the path to the plan file, such as plans/plan.md, for architecture evaluation. The system ingests this document to perform cross-file linking and generate a formal evaluation report.