Claude Code 可魔改：用 Jev 做代码审查层
Anthropic 让 Claude Code 支持自定义，作者提议加一层 Jev 式 System One 审查，扫架构违规、bug 和规格漂移。
You can basically build your OWN Claude Code now... Anthropic just made Claude Code moddable. this is going to get ridiculous. some of my favorite ideas: > add a System One code-review layer like Jev to scan the repo for architecture violations, bugs, spec drift + risky changes > turn Claude Code into a DeepSeek-style harness with builder → reviewer → verifier loops > make it Hermes-like with persistent specialist agents + shared context > add architecture / security / test gates before code gets accepted > build your own agent dashboard directly into the UI > swap out tools + features you don't like > build completely custom workflows around your own codebase I'm 100% going to make a completely unhinged Claude Code setup that can call ChatGPT models directly from the UI too lol.
这条是引用着另一条写的：x.com/claudedevs/status/2105721434807083061