Browser-native deck
Brutal Honesty Kit
Evidence-bound claims for AI-assisted delivery.
The failure mode
Generated confidence can outrun verification.
Scaffolds, mocks, missing smoke runs, and broad readiness claims become expensive when they enter review as finished work.
The convention
Claims must match evidence.
- Declare what changed.
- Name what was actually tested.
- Keep caveats visible.
- Carry unresolved debt forward.
The kit
Repo-native enforcement.
Rulebook, templates, validators, smoke gates, and review artifacts make honest status cheaper than narrative cleanup.
Adoption path
Start with PR structure. Add CI when the team is ready.
The package is static, portable, and designed to live beside the code it governs.