Comment “FABLE” for the prompt I used to do this

Five things you can do with the new Fable 5.2 that'll make you better than 99% of people with Clawed code. Do a massive refactor of your entire code base. One thing you might have done with your entire code base is create rules and skills that clutter your context which makes your agent extra slow because you haven't taken into account progressive disclosure in your systems. There

are three ways an agent's going to find something. Either it's always loaded like your clawed.md, it's loaded when a folder is opened, or it's found by search. When you tell clawed things like from now on do this from now on do that and you don't have progressive disclosure rules inside of your system it will cram everything into the first stage which means that all of the skills that

you're not even using right now will all get loaded into context which will just inflate the call itself the refactor took all of my skills and divided them into specific departments these are all of the marketing and content skills that i have which will now only get loaded when i need to do marketing stuff so now my claude should run way faster than it was beforehand because it would

take a minute and a half sometimes for it to get going but now it should only take like 30 seconds Run insights and do every plan. Insights is a slash command that opens up this artifact, which reads through all of the past three months of conversations that you've had. It tells you where things are currently going wrong for you and then suggests additions to

your Claude.md, new skills and hooks and ambitious features for you to try as well. Feed the entire report back to Claude and do all of the things that it recommends and then do a second wave with all the things that are on the horizon as well. Ask Fable to audit your conflicting files That refactor highlighted elements of my Claude that were different elements that are inside of my skills and inside of my second brain Based on this concept of progressive disclosure if I look at one and then claude finds all three and now has conflicting information and then ask me a bunch

of stupid questions. So part of the refactor and part of insights is also finding some of these conflicts so that you can get some answers once and for all about how you're going to do things. Which takes us neatly onto the next one. Loosen all of the hard rules that are across your repo. Claude told me that I had about 3,700 of these that I know comes from me beating Claude over the side of the head to make sure things are going correctly. But most of these are inside of

skills, which is fine because skills are meant to be protocols that are strictly followed. However, they are indications of where your systems are weakest. And for me, I know that those are mostly the video skills. Where they concentrate are indicative of where either you are weak or where Claude in general is weak. So sometimes it's a good idea to look at other things

that can help you with that. New GitHub repos like hyperframes, new skills, or just getting better at understanding what the hell you're trying to do. As long as these conflicting messages are concentrated in your skills as opposed to your claw.md, you should be fine. If it's inside your claw.md, big no-no. Get these fixed. Give Fable a really, really big task and

give it user stories. Don't tell it how to do the work. User stories describe the experience of what it's like to have the solution that you're asking for. So if you're building, for example, a new automation, you may want the user story to say something like, when I click this button, I should see this and this appear in front of me fable is smart and wise enough that you don't have

to tell it how to get to the end point you could just tell it what the end point needs to look like in as much detail as humanly possible fable will figure it out comment fable and i'll send you the prompt that i used to get all of this done