No Trace, No Grade: A Bootcamp Lab on Agent Flight Recorders

Описание и применение

Лабораторная работа в кодинг‑буткемпе, где студенты должны предоставить лог сессии и diff, чтобы преподаватели могли воспроизвести действия агента.

Задача

Students cannot reliably demonstrate what their AI agent did during a lab, making grading difficult and subjective.

Результат

The lab enforces reproducible, machine‑readable evidence of agent activity, improving grading reliability and peer review.

Рабочий процесс

• Set up a unique workspace directory and initialize a Node project
• Create the broken production file, tests, and recorder script
• Run the tests once to confirm failure
• Use the recorder wrapper to execute the test suite while logging all actions
• Review the generated session.jsonl and git diff to ensure completeness
• Submit the log and diff as the lab deliverable

Доказательства

Not a prompt dump. Not a PNG. A JSONL log plus a git diff that a classmate can follow without asking you what "it" meant.

Источник

This wrapper does not call a model. It records what you (or your agent) did.

Источник