Task 278: Science Accuracy Report External Review

Status: DONE

Motivation

The science-accuracy report (reports/data/summary/science-accuracy.md) summarizes quantitative verification of scientific claims across all 5 episodes, checking depicted values against real-world data. Following the established review pattern (Tasks 266-277), conduct an external review to check data integrity, readability, factual consistency, and accessibility to newcomers unfamiliar with the series or orbital mechanics.

Issues Found and Fixed

External review (Sonnet agent) found 12 issues:

1. HIGH: Navigation error formula used wrong distance — total solar distance 14.72 AU (2.202×10⁹ km) instead of remaining distance to Uranus 4.48 AU (6.704×10⁸ km). Formula also used sin instead of tan. Result (14,393,613 km) was correct but intermediate calculation was wrong.
2. HIGH: EP02 shortening factor was 8× — corrected to 42× (Hohmann 3,671d ÷ trim-thrust 87d).
3. MEDIUM: No newcomer introduction — added "はじめに" section with series overview and navigation links.
4. MEDIUM: Oberth effect 3% vs 0.07% gap under-explained — expanded reference note to clarify classical Oberth (0.07%) vs mission-level composite (3%).
5. MEDIUM: EP4/EP5 ΔV table ambiguity — added clarifying notes that EP4 and EP5 share same route, EP5 uses composite.
6. MEDIUM: RK4 rows labeled as "depicted" when they're computational verification — prefixed with 【計算手法検証】 and clarified depicted values.
7. MEDIUM: EP02 trim-thrust arrival velocity inconsistency (v∞≈90 km/s vs 0.61 km/s capture) not mentioned — added to 総合評価 as known open issue.
8. MEDIUM: EP4 "8.3日" figure is hypothetical brachistochrone, not depicted — added "(300t仮想Brachistochrone)" annotation.
9. LOW: vis-viva equation entries had no hyperlinks — added Wikipedia links.
10. LOW: No navigation links to episode reports — added links throughout section headers and body.
11. LOW: Glossary missing Brachistochrone, RK4, ICRP, プラズモイド — added 4 terms.
12. LOW: Peak vs average velocity conflation in cruise speed 3000 km/s entry — noted but left as-is (93% accuracy with approximate status is defensible).

+6 regression tests (nav error remaining distance, EP02 shortening factor, newcomer intro, EP02 velocity issue, episode links, glossary completeness). Tech-overview stats refreshed (278 tasks, 2,083 TS tests, 2,674 total). Total: 2,083 TS tests.