Free: 10-item self-check for Claude Code autonomous ops

[yurukusa]

Running Claude Code autonomously? Here's a quick self-check I put together from 200+ hours of autonomous operation.

10 items covering the gaps that caused real failures:

1. Context monitoring (session died at 3% — 1 hour to recover)
2. State saving on session end
3. Blocking dangerous commands (rm -rf, git reset --hard)
4. Git backups before changes
5. Syntax checking after edits
6. Rate-limiting AI questions to the user
7. Pre-send checks for external actions (X posts, PR creation)
8. Root cause analysis instead of brute-force fixing
9. Activity logging
10. Operational rules in CLAUDE.md

Each item includes what can go wrong and a one-line hook implementation.

Full checklist (Gist): https://gist.github.com/yurukusa/23b172374e2e32bdff7d85d21e0f19a2

Items #1-#5 are implemented as ready-to-use hooks in this repo (install.sh → 15 min setup).

MIT licensed. No signup, no paywall.