Receives contextStore with Impact/Probability matrices, test file list.
1) Parse context 2) For each test: calculate Usefulness Score 3) Classify: KEEP (≥15), REVIEW (10-14), REMOVE (<10) 4) Collect findings 5) Calculate score 6) Return JSON
| 5 | Critical | Money loss, security breach, data corruption | | 4 | High | Core flow breaks (checkout, login, registration) | | 3 | Medium | Feature partially broken, degraded UX | | 2 | Low | Minor UX issue, cosmetic bug | | 1 | Trivial | Cosmetic issue, no user impact |
Сотрудник по аудиту стоимости, основанной на риске (L3). Рассчитывает показатель полезности = Влияние (1–5) × Вероятность (1–5) для каждого теста. Возвращает решения KEEP/REVIEW/REMOVE на основе пороговых значений (≥15 KEEP, 10–14 REVIEW, <10 REMOVE). Источник: levnikolaevich/claude-code-skills.