When an agent says done, which files did it test?

Описание и применение

Squadra позволяет отслеживать, какие файлы проверяли агенты, сохраняя доказательства и обзоры в локальном реестре, что повышает надёжность слияния кода.

Задача

When multiple coding‑agent windows work on the same project, the chat messages that tests passed or a review completed do not indicate which files were actually checked, making it hard to trust the results before merging.

Результат

Squadra enables traceable coordination of tests and reviews across multiple agents, ensuring that the final merge is based on current, verified evidence.

Рабочий процесс

• 1. Author agent (Claude) modifies code to fix a bug.
• 2. Agent runs agreed test command; test passes on the candidate files.
• 3. Reviewer agent (Codex) reviews the frozen candidate and records approval.
• 4. Someone edits the checked files, changing the candidate.
• 5. Tests and review are repeated for the new candidate; a final check ensures the candidate, proof and review match before closure.

Доказательства

Squadra keeps assignments, acknowledgments, proof, reviews and decisions in a shared local registry.

Источник

Its prove command records the check and file fingerprints.

Источник

Before a task can close, the candidate, proof and review are checked together again.

Источник

The toolkit requires Python 3.12+ and Git; Node.js is needed for the dashboard checks in the self‑test.

Источник