review-result-schema-is-normalized

IN premise

Every result dict returned by `parse_review_response` is guaranteed to have exactly six keys (`id`, `valid`, `sufficient`, `necessary`, `unnecessary_antecedents`, `comment`) regardless of what the LLM returned, via normalization with safe defaults.

Summary

The review system's response parser always produces a consistent, predictable structure with six fields, even if the LLM returns incomplete or malformed output. This means downstream code can safely access any of those fields without checking for missing keys, simplifying error handling throughout the review pipeline.

Dependents

These beliefs depend on this one:

Details

Sourceentries/2026/05/05/reasons_lib-review.md