AKUNUR 3.0.0-alpha.5 VALIDATION REPORT

Automated checks completed:
- All JavaScript files pass Node syntax checking.
- Core smoke test passes.
- Full-history and lightweight execution remain equivalent over 150 deterministic decisions.
- Search previews do not mutate source state.
- Undo restoration passes.
- Decisive-corner and non-winning Echo-placement regressions pass.
- Canonical keys, threefold tracking and no-progress tracking pass.
- Tactical agent returns a legal action without mutating the source state.
- Tactical agent selects an available immediate third-corner win.
- Tactical agent selects a legal mandatory Echo Piece placement.
- Browser-style core, analysis and all three agent bindings pass.

Representative headless validation on the build machine:
- Tactical as White defeated Random in 5/5 games, averaging 102.6 decisions.
- Tactical as Black defeated Random in 5/5 games, averaging 125.6 decisions.
- Tactical defeated Legacy in both completed White-side samples (73 and 77 decisions).
- Tactical defeated Legacy in both completed Black-side samples (80 and 76 decisions).
- In a two-game Tactical-v-Tactical sample with a 600-decision / 120-no-progress boundary, one game ended in a genuine 3-1 corner win and one ended as a 2-2 analytical no-progress draw.
- Result accounting remained valid and no terminal-state consistency issues were reported.

These are development smoke samples, not statistically sufficient balance evidence.

Manual checks still required:
- Load web/index.html through a static server.
- Confirm board appearance and interactions match alpha.4/2.10.5.
- Confirm the AI control cycles Tactical -> Legacy heuristic -> Random.
- Run Tactical autoplay and inspect whether corner plans appear purposeful.
- Confirm analytical autoplay stopping leaves the position manually playable.
- Exercise corner claim, Echo placement, Echo Point use, capture and undo.
