Interesting that you started with commands rather than hooks. I went the other way round and found hooks were the bigger unlock. Once your Stop hook runs the test suite automatically, you stop having to tell the agent to test things.
If you're comparing Claude Code to other tools, Codex CLI just shipped its own hook system (SessionStart and Stop events). Way simpler config but the same principle: deterministic shell commands that fire regardless of what the AI decides. I wrote up the setup and comparison here https://reading.sh/codex-cli-has-hooks-now-stop-stuffing-agents-md-c181465fe271
Curious what made you stick with Copilot over Claude Code in the end? I've been bouncing between Claude Code and Codex CLI depending on the task.
Interesting that you started with commands rather than hooks. I went the other way round and found hooks were the bigger unlock. Once your Stop hook runs the test suite automatically, you stop having to tell the agent to test things.
If you're comparing Claude Code to other tools, Codex CLI just shipped its own hook system (SessionStart and Stop events). Way simpler config but the same principle: deterministic shell commands that fire regardless of what the AI decides. I wrote up the setup and comparison here https://reading.sh/codex-cli-has-hooks-now-stop-stuffing-agents-md-c181465fe271
Curious what made you stick with Copilot over Claude Code in the end? I've been bouncing between Claude Code and Codex CLI depending on the task.
Anthropic really does make the best products right now in my opinion.