Loading page...Загружаем страницу...正在加载页面...
AtlasRepo

I counted every skill my AI invoked across 55 sessions. Here’s the leaderboard.

Описание и применение

Автор подсчитал, сколько раз каждый AI‑skill использовался в 55 сессиях разработки ParkEase, выявив, что роутер потоков, навыки дизайна и проверки безопасности были наиболее активными, а тест‑ориентированная разработка, хотя и вызывалась редко, оказала значительное влияние.

Задача

Determine which installed AI skills actually contributed most during development of ParkEase.

Результат

The workflow router (‘flow’) was invoked 28 times, the most of any skill, followed by design skills and safety nets. Test‑driven development, though invoked only 4 times, produced 3,246 unit tests and 1,080 integration tests, showing high impact. The study informs which skills to prioritize when starting a new project.

Рабочий процесс

  • 1. Gathered all 55 ParkEase sessions (10 Sep–19 Oct).
  • 2. Searched Claude Code transcripts for explicit skill invocations.
  • 3. Counted invocations per skill.
  • 4. Compiled leaderboard table.
  • 5. Analyzed surprising patterns and impact.
  • 6. Listed recommended skills for future projects.

Доказательства

I searched all 55 ParkEase sessions (10 September to 19 October) for every time the agent explicitly invoked a skill.

Источник

The router came first. <code>flow</code> doesn't write a line of code. It decides which other skills run.

Источник

Only 4 explicit calls, because it was loaded once per build phase and stayed loaded. Its output, 3,246 unit tests and 1,080 integration tests, is the clearest proof that invocation counts undersell the skills that stick.

Источник

  • Дата записи: 2026-11-01T07:59:46.787Z