Consistency in execution across sessions.

[appressman]

I'm leveraging PAI (mine is named Eve) to try to leverage the overarching facility of maintaining context throughout sessions but having a struggle. In something like a Claude code guided construction of manual workflows, in my case inside the HighLevel CRM, where I'm manually creating what Eve has created a playbook to do, I find that as I close down a day's work and resume the next in another day's session things get forgotten. For instance I'll be handed a list of tags to that include an entry like "webinar-attended" and if I stop that session before completing it, the next time I open up Eve and ask for that list it is likely to call that tag "attended-webinar". It's as if with each session Eve forgets what she's done (I have the todo and commits done in each session) or what she can connect to or other capabilities. In other words I'm trying to find a way to "code before prompt" or otherwise make the AI work less in a predictive and more in a regimented mode. Perhaps there's some command to update the contexts, history etc that I'm just not aware of.

Any help much appreciated.

[danielmiessler]

I would check out the history system and make sure you're logging all of those sessions and/or manually recording them into lessons learned within the history system.

[appressman]

I'm going to have to have "Eve" explore the docs and determine exactly how to ensure the logging. Is there a command for manually recording sessions into lessons learned?

Thanks and great job! Incredibly useful even with the extra effort require to keep Eve on task.