Round 5 — Explore Science review of v6

Zero major, thirteen minor issues, 42 merits. The consequential ones: the dynamic interiority filtering inside permutations was ambiguous (B1 — verified, plus a reporting omission corrected); no format-matched dummy control exists for the persona-prefix contrast (B2 — construct confound conceded); and the exact episode-level family gate is severely underpowered at six episodes per arm (B3 — adopted, tested, and turned into the p=0.075 attainability result). Figure 5's misattribution (C1) was corrected. The full disposition matrix and validation record are archived beside it.