Problem
Software teams increasingly rely on agents that can produce code, tests, summaries, and release narratives in the same work session. That speed changes the review problem. The reviewer must now distinguish between implemented behavior, generated scaffolding, local demonstrations, and unsupported confidence.
Principle
Every completion claim should be bound to evidence with a known scope. When evidence is weak, the claim should remain narrow. When uncertainty survives a cycle, it should be carried forward instead of being softened into normal backlog language.
Implementation
Brutal Honesty Kit installs a convention into the repository itself. It pairs rulebook language with templates, closed-set fields, validators, smoke commands, and reviewer expectations. The result is not a replacement for judgment; it is a way to make judgment possible before the narrative hardens.
Adoption
Teams can start with PR-body structure and local validation, then expand into CI checks and formal fresh-agent review. The highest-value change is cultural but the enforcement surface stays practical: files, commands, and artifacts that live beside the code.