pipeline-plan-reviewer

Plan review involves automating rigorous evaluation of a plan against architecture patterns and acceptance criteria to prevent gaps during delivery. It validates AC mappings, architecture alignment, scope completeness, dependencies, tests, and design coverage, outputting a structured plan-review.md with a verdict.

Consensus-driven plan review is a multi-agent assessment process that evaluates a project plan against architecture patterns and acceptance criteria. It runs a 3x3 assessment for medium-to-large complexities to ensure handoff readiness and prevent gaps early in delivery.

To prepare inputs for an architecture plan review, provide a prepared plan.md along with required input contracts including artifact_path, key_decisions, known_risks, complexity, and tech_stack_adapter. You can also include an optional design_adapter to guide the evaluation.

Yes, you can use plan review for medium-to-large complexity software projects. The consensus-based review specifically runs a multi-agent 3x3 assessment tailored for medium-to-large complexities to validate scope completeness, dependencies, tests, and design coverage.

A plan-review.md output contains a structured verdict and iteration data for coder and reviewers. It captures the results of validating acceptance criteria mappings, architecture alignment, scope completeness, dependencies, tests, and design coverage to ensure handoff readiness.

You need to run a plan review in Phase 2 of the software delivery pipeline, after feeding the prepared plan.md and inputs to the reviewer. This prevents gaps early in delivery by ensuring design and architecture alignment before handing off to coders.