0.33.0-alpha · core 2026-06-22

Estimation bias correction (global criterion) — New rules across cali-product-shape-up, cali-product-tech-planning, and cali-product-plan-critique to counter model overestimation bias: scope count warnings are informational, cuts_needed must be based on value overlap not perceived complexity, final decision is always human.
E2E-first testing priority — cali-product-testing-ai-code and cali-product-testing-execution reordered to prioritize E2E/behavior tests over unit tests across all appetite levels (Lean=1 E2E happy path, Core=E2E+variations, Complete=full E2E coverage).
Estimation is relative, not absolute — plan-critique feasibility checklist now uses relative comparison levels (Low/Medium/High) for scope ranking, never absolute numbers.
Scope adjustment bias note — Shape Up scope adjustment warns when model recommends removing items due to perceived complexity.

The mechanical warnings above (scope count, spec lines) are indicators, not gates.

cali-product-testing-execution phases inverted: E2E Browser Testing (Phase 1) → UI Quality (Phase 2) → Unit Tests (Phase 3) → Code Review (Phase 4) → Final Checklist (Phase 5). Decision tree, examples, edge cases updated.
test-behavior scopes now mandatory in all appetites (was Complete-only).
All new skill text in English (translated from Portuguese).
Updated appetite-consistency.test.ts to match English assertion.