I’ve developed a prompt for claude which it has never been fully able to follow. I’ve asked claude to help me debug and refine the prompt, but after 40+ tries, I see that it simply can’t do anything to resolve this. So I joined Make in hopes that this will help me nudge Claude along.
Without getting into the actual prompt itself (which is about 2,000 words), can someone steer me on creating a scenario which somehow, repeatedly remind Claude to refer back to the prompt?
The project I’m working on requires that Claude produce long textual documents (up to 40,000+ words).
Though my prompt gives it instructions to abide by certain style (and other rules), it tends to forget those rules as it proceeds to generate the text.
It also is unable to proceed very far without asking me for permission to continue, despite carefully written instructions not to pause or stop.
To compound this, because I have to break the task into multiple chats, and because it’s essential that it remembers what it generated in the prior chat, I need to figure out a way to incorporate the last chat into the documentation (or prompt) for the new chat.
So I’m hoping that Make can help out. If it can simply (a) monitor Claude’s output, (b) tell it to continue, or tell it to open a new chat and continue, and (c) always refer back to the prompt before writing the next section of txt, that would be enough.
Thanks!