A true master who uses hidden commands like Claude Code and Codex every day ‼️ I thoroughly organized it today (at the collection level) Claude Code ： 1./btw Suitable for temporary inquiries. When the main task is running and you suddenly want to ask a small question, use this to avoid disrupting the current rhythm. 2./rewind Retract the artifact. Wrote incorrectly, made changes, and was not satisfied with the experimental plan, directly returned and started over. Dare to try and make mistakes is the true efficiency. 3./model opusplan This is crucial. Strong models are responsible for planning and thinking, while fast models are responsible for execution and implementation. Deep when necessary, fast when necessary. 4./simplify The fact that code can run does not mean that the code is clean. This command is more suitable for organizing, reducing redundancy, and organizing structures. 5./fork Keep one idea for now, and then open a new branch to try other solutions. It is very useful for architecture, implementation, and running different routes. 6./remote-control Cross device continuation tasks are very useful. You don't have to keep the current computer, you can continue with the current session. 7.Shift + Tab Switching modes is very convenient. Execute and plan these states easily without constantly reorganizing the process. 8.Option + P / Alt + P Quickly cut the model without clearing the current input. Halfway through writing, I suddenly want to change my mind, which saves a lot of time. Codex： 1./model Cut the model. Strong models for complex tasks and light models for simple tasks, don't let all tasks follow the same configuration. 2./permissions This is very important. Codex can read files, modify files, and run commands, so the permission boundaries must be well controlled. You give it as much authority as it can do. 3./compact Long conversations are essential. When the context becomes longer and longer, compress it and keep the focus, and continuing to work later will be much lighter. 4./fork Keep the current route and try another repair method on a new route. It is very easy to compare schemes. 5./review Very useful before submission. Not to rewrite it, but to help you identify risks, problems, and blind spots from a code review perspective. 6./diff Look directly at what has been changed. Don't rely on your brain to recall, just look at the differences, it's the most stable. 7./agent Suitable for multi-agent threading scenarios. Different threads do their own things, you can always switch over and continue pushing forward. 8./fast Small task acceleration is very useful. When there is no need for heavy reasoning, simply accelerating the pace will greatly enhance the sense of progress. 9. @ File Pull the file directly into the context. It is much more efficient than manually describing file paths. 10.! command Run the local shell directly. For those who truly write code, this type of ability is more important than just being able to say a few more words. 11.Enter / Tab / Esc This set of quick logic is very valuable. During execution, instructions can be inserted, the next step can be scheduled, and the fork can continue to branch back to the previous round, making the entire interaction more like you are driving an agent in real-time.