Hi HN — I built alive-analysis, an open-source workflow kit that turns AI coding agents into structured data analysis partners.

Problem: AI-assisted analyses are often throwaway chats. A month later, you can’t trace why you reached a conclusion.

Solution: It enforces a 5-stage loop (ASK → LOOK → INVESTIGATE → VOICE → EVOLVE) with checklists, and saves analyses as Git-tracked markdown files. Quick mode: 1 file. Full mode: 5 files.

Works in Claude Code and Cursor.

I’d love feedback on:

1. Does the ALIVE loop match how you do investigations / experiment reviews?

2. Which checklist items feel missing or unnecessary?

3. What would make this usable in a team setting?