What did your coding agent do last month?

Описание и применение

Сканер agentdryrun позволяет без выполнения действий просмотреть и классифицировать команды, выполненные агентом Claude Code, выявляя потенциально опасные операции.

Задача

Users want to audit the actions taken by Claude Code’s coding agent without executing or altering anything.

Результат

The scanner identified 48 actions, 24 would have passed, 19 would have needed approval, and 5 would have been refused, revealing real force pushes to main.

Рабочий процесс

• 1. Clone the agentdryrun repo
• 2. Run python -m agentdryrun scan --no-previews to scan the transcript
• 3. Review the generated report card
• 4. Adjust the JSON rules file as needed
• 5. Re‑run to refine results

Доказательства

Claude Code already keeps a transcript of every session on your machine, in ~/.claude/projects

Источник

python -m agentdryrun scan reads those files, names what the agent attempted, and judges each action against a rules file as a dry run

Источник

A sample report card. Every command in it is invented.

Источник

It matches patterns in commands. It misses a script that deletes files, python -c, and paths built at run time, and it flags a few harmless things.

Источник