Nolen began their writing career in 2019, with three years dedicated to editing the Creative section at MakeUseOf. Their expertise lies at the crossroads of technology and creativity, covering areas like photography, video editing, and graphic design.
Outside of work, you'll often find Nolen diving into a good book, writing their own stories, or playing video games.
Claude Code sessions have a shelf life. Once one fills up with context and starts dragging, the sensible move feels like clearing it or opening a new session. Except, either way you're left with a session that doesn't have a clue what you were in the middle of. And then you end up spending the first ten minutes catching it up with what you were working on and re-explaining some concepts from a to-do list.
This is exactly what the CLAUDE.md is supposed to fix, but the catch is that it only knows what you remember to write in it, and much of my Claude Code use never got written down anywhere because most of my session history isn't that quintessential to future projects. Until it is, then I'm left scrambling. For ages, my fix was skills and the CLAUDE.md, since plugins seemed like too much fuss, but then I actually started using them and regret waiting this long.
Claude already remembers, just not the right things
What carries over between sessions isn't always your call
Claude Code does carry some things between sessions, to be fair. Each session in the desktop app keeps its own chat history, but a new one starts with a fresh context window, and two things load into it at the start: a CLAUDE.md file you write yourself, and auto memory, which are notes Claude decides to keep on its own.
Auto memory is the one that's going to throw off your projects because it's basically just Claude picking what it deems important enough to be worth saving, and Anthropic's docs even say it doesn't save something every session. Those auto-memory notes don't even live in the folder you're working in, and instead Claude saves them to a .claude folder in your OS user profile - so you'll see something like "saved 2 memories" without knowing what actually made the cut.
CLAUDE.md should have covered it, but the thing is, it's not exactly built for memory. Anthropic says to keep it to standing rules Claude needs in every session, and under about 200 lines, since the whole file loads each time and longer ones get followed less reliably. So the state I've left a project in yesterday doesn't necessarily belong in this file.
So, I turned to plugins. A plugin is a package that bundles extra skills and commands, sometimes with connections to other apps - you can install plugins pretty easily through both the terminal or the desktop app. Because a plugin can add its own commands, it can handle the saving part for you, so instead of editing CLAUDE.md by hand or hoping auto memory caught something, you run a command and it writes down where things stand in a file you can actually open and read. And I ended up installing two that changed my Claude sessions.
The productivity plugin that keeps track of where everything stands
A plain task file that asks questions instead of guessing
Productivity is one of Anthropic's own plugins, built more for general work than for code, which is exactly why it suited my workflow so well since it's mostly computer admin. It handles tasks and memory. Running /productivity:start creates a TASKS.md file in your folder, split into Active, Waiting On, Someday, and Done, plus a small dashboard page you can open in a browser.
You can point it at an existing to-do file or paste a list in, which is what I did, and it sorted everything itself, even turning a loose "before Friday" into an actual date. Updating it after that simply means telling Claude what happened in normal words. For example, I said one task was stuck waiting on feedback and it moved it over. It can pull tasks from apps like Asana or Linear too if you want to set up those connections.
Something I like about this plugin is that it won't try to guess your shorthand. For example, half my tasks are abbreviations, and it actually asked me what each one meant and which project it belonged to. Those answers got saved as plain files in the folder with a CLAUDE.md for quick reference and a memory folder for longer notes. When I open a new session in that folder now, /productivity:update runs through all the tasks, including how many days are left on anything due.
claude-md-management turns one-off corrections into rules
The file I never updated now gets updated at the end of every session
The second plugin is claude-md-management, also from Anthropic, and it's tiny. There's one command called /revise-claude-md and one skill called claude-md-improver, and both are about keeping your CLAUDE.md up to date. I think this one was pretty clearly made for projects centered around coding, but nothing stops you from using it for other tasks.
Using it is pretty simple. You have your sessions as per usual, and if you told Claude something it should know next time, for example a naming rule for new folders, you run /revise-claude-md before closing the session. It will look over everything that happened in the session and draft the additions as a diff. Something won't actually get written until you approve it, which solves the issue of auto-memory picking for you.
The Improver command basically grades the whole file. Mine came back at 79 out of 100, meaning it caught things I didn't think to write down, for example a few folder names with spaces in them that can trip up commands. One thing to keep an eye on is that /revise-claude-md suggests new lines but never removes old ones, so the file keeps growing. Running the improver now and then helps trim it back.
I'm ending my sessions properly now
Oddly, the biggest change has been how I end sessions rather than how I start them. Ending a session used to mean just closing the window, but now I take a second to tell Claude what's been done and approve the new rules. It's a small habit that can spare you from potentially wasting hours every month re-explaining yourself.
