Show: Bash hooks for autonomous Claude Code sessions (free, MIT)

[yurukusa]

I've been running Claude Code autonomously for 200+ hours (built a 15K-line game this way) and extracted the core hooks into an open-source starter kit:

https://github.com/yurukusa/claude-code-ops-starter

4 hooks:

• Context monitor: Staged warnings at configurable thresholds, auto-checkpoint at critical
• No-ask-human: Blocks AskUserQuestion so the AI decides + logs to pending_for_human.md
• Syntax check: Auto-verifies Python/Shell/JSON after every Edit/Write
• Decision warn: Flags destructive commands (rm -rf, git reset --hard, etc.)

One-command install, zero dependencies, MIT license. Hooks degrade gracefully without optional deps (jq, python).

git clone https://github.com/yurukusa/claude-code-ops-starter.git
cd claude-code-ops-starter
bash install.sh

Would love feedback from other Claude Code users. What hooks do you run in your setup?