coherence-reviewer

To check a design document for internal contradictions, you need to analyze statements across sections like scope and requirements to find conflicting claims. This process surfaces contradictions, terminology drift, and broken references that could cause implementers to diverge.

Terminology drift in technical planning documents occurs when the same concept is named or defined inconsistently across different sections. Detecting this drift requires analyzing the document to identify instances where identical concepts are referred to by varying names or conflicting definitions.

Detecting unresolved dependencies in architecture proposals involves scanning the text for dependencies mentioned without resolution. The review highlights these gaps alongside conditional logic gaps and unclear responsibilities, providing confidence levels and suggested structural groupings.

Finding ambiguous quantifiers in requirements documents is achievable by applying an automated review that highlights quantifiers without bounds. The analysis flags these ambiguities alongside broken internal references and forward references, suggesting logical groupings while preserving original IDs.

The best way to review a technical plan for structural issues is to flag forward references and ungrouped multi-concern requirement lists. This approach suggests logical groupings while preserving original IDs, ensuring a prioritized list of coherence findings with confidence estimates.

Coherence review works effectively for merge requests and PRs by analyzing attached planning documents. Running during design reviews or PRs produces a prioritized list of coherence findings, including contradictory statements and unresolved dependency assumptions with confidence levels.