AKUNUR 3.0.0-alpha.7 VALIDATION REPORT

Automated checks completed:
- JavaScript syntax checks across core, four agents, Lab and browser files.
- Core smoke test and full/lightweight execution parity.
- Undo restoration and search-preview isolation.
- Immediate decisive-corner ending and non-winning Echo placement regression.
- Mutually exclusive outcome accounting.
- Canonical position, repetition and strategic-plateau diagnostics.
- Capture-outcome and capture-relationship classification.
- Productive versus non-productive capture accounting.
- Tactical v1 regression suite unchanged and passing.
- Strategic legal/pure/seeded selection, immediate win, reversible capture discipline, forced retarget and agent-memory round-trip tests.
- Coordinate-neutral target scores and 180-degree rotation symmetry test.
- Seeded whole-game reproducibility.
- Replay reconstruction, continuation and Strategic memory restoration.
- Browser-style module and four-agent binding.

Representative behavioural checks:
- Strategic self-play produced genuine wins for both colours in short smoke samples.
- A Strategic player can record and execute a forced target change after a plan-relative plateau.
- Reversible capture relationships are reported as non-productive unless the same action also creates material, corner or formation progress.
- Strategic remains slower per decision than Tactical because it evaluates the Tactical baseline and an additional strategic intervention layer; current performance remains suitable for local analysis.

Manual checks still recommended after upload:
- Load the web build and verify the alpha.7 header and cache refresh.
- Cycle Strategic, Tactical v1, Legacy and Random modes.
- Run browser autoplay through a normal win and through a strategic-plateau stop.
- Exercise first corner token capture, Echo placement, Echo Point movement/capture, corner recapture and undo.
- Continue one alpha.7 replay once with `memory` and once with `reset` and confirm the mode is reported.

Known validation limitation:
- The two exact alpha.6 plateau replay JSON files used on Chris's laptop were not present in the build workspace, so alpha.7 includes synthetic plateau/capture regressions rather than embedding those exact positions. The finished Lab can load those original replay files for direct Strategic-versus-Tactical or Strategic self-play comparison.
