The one prompt that installs all five
This is the whole point of the page. Open Claude Code, paste the prompt below, and it installs four of the five for you and reminds you to type the fifth. Each one gets a plain-English report as it goes, so you always know what just happened on your computer.
Before you paste
You need Claude Code open — Mac, Windows or Linux all work. Any folder is fine, but the project you use most is best. It takes about 15 minutes, and Claude will ask permission before it installs each thing. Say yes. If anything fails it tells you why and keeps going, so you never end up half-done and confused.
Then type this inside Claude Code
Anthropic's official plugin has to be installed by you, not by Claude. One line, hit enter, done.
Only want one or two?
Paste the big prompt anyway and delete the numbered steps you don't want before you hit enter. Claude only installs what's left.
Which one fixes your problem
Claude Code has four habits that make people give up on it: it runs out of usage, it eats tokens on junk, it forgets everything when you close it, and it starts every project with a blank setup. Each of these fixes one of those. Find your annoyance below.
"I don't know what to set up"
Claude Code Setup. Anthropic's own plugin reads your project and tells you which hooks, skills and MCPs would help. Install this first, whatever else you do.
"It forgets everything when I close it"
Claude-Mem. Remembers what happened last session and brings the useful bits back. Install when you're sick of re-explaining your business.
"Every message feels expensive"
Headroom. Squeezes the bulky stuff (logs, file dumps) before Claude reads it. Same answers, fewer tokens. Install when you work with big files or long sessions.
"I keep hitting the usage limit"
OmniRoute. When your Claude quota runs out it hands the job to another model so you keep working. Install when you've seen the limit message twice this week.
"My skills never get better"
Task Observer. Watches how you actually work and suggests updates to your skills. Install once you've got a handful of skills and they're not keeping up.
The prompt installs all five. If you'd rather go one at a time, that's the order I'd do it in: Setup first, then memory, then Headroom, then OmniRoute the day you hit a wall, and Task Observer last.
The five, one at a time
What each one actually does, what changes once it's in, a prompt to prove it's working, and the catch. Strictly speaking two of these are plugins, one's a skill and two are little helpers that run on your own computer. I'm calling them all plugins because that's what they feel like.
Claude Code Setup · Anthropic's official setup check
This goes first because it's made by the people who make Claude Code. Ask it "help me set up Claude Code" and it reads your project and comes back with the top one or two things worth adding in each category: MCP servers, skills, hooks, sub-agents and slash commands. It never touches a file. It only recommends, so nothing can break.
After install, nothing looks different. The change is that you can finally ask Claude what to set up instead of guessing.
The catch: it reads, it doesn't do. You still say yes to each recommendation and Claude sets it up in a separate step. That's a feature, not a bug. See the plugin on GitHub →
Claude-Mem · memory between sessions
Close Claude Code and it forgets your project, your preferences and the thing you fixed yesterday. Claude-Mem records what happened in each session, compresses it, and drops the relevant bits into your next session before you've typed a word. It also adds a search skill, so you can ask "what did we decide about pricing last week?" and get a real answer.
After install, restart Claude Code. Context from earlier sessions starts appearing on its own. There's no manual step.
The catch: it has just renamed itself Grok Mem. The package is still called claude-mem and the install is the same, so don't panic when you see the new name. The hosted memory service is free for 30 days, then it falls back to running on your normal Anthropic plan, which still works and costs nothing extra. Wrap anything sensitive in <private> tags and it won't be stored. See it on GitHub →
Headroom · fewer tokens per message
Claude Code reads a lot of junk on your behalf: log files, giant file dumps, tool output. Headroom sits in the middle and squeezes that down before Claude sees it. The makers quote about 20% fewer tokens on everyday coding work and far more on messy data files. Everything runs on your own computer. Nothing is sent anywhere to be compressed.
After install there's one change to your routine. When you open Terminal, you type the line below instead of claude. That's it.
Then, inside that session, check it's earning its keep:
The catch: it only works in a session you started with that command, so make it a habit. Intel Macs need the Docker version instead (the installer prompt checks for you). It also installs a code-navigation helper called Serena alongside it, which is fine, and the prompt switches off its anonymous usage ping. See it on GitHub →
OmniRoute · keeps you working past the usage limit
Your Claude plan has a usage cap, and it always runs out mid-task. OmniRoute is a free switchboard that connects Claude Code to 350-odd other AI providers, including a big stack of free ones. When your Claude usage runs out, it hands the job to the next best model instead of stopping you. The makers put the free allowance at around 1.5 billion tokens a month, which is their number, not mine.
The installer prompt only gets it running. To actually switch Claude Code over, use the prompt below the day you hit a limit. It's reversible.
The catch: when it falls back, a different model answers. Not Claude. For everyday admin that's fine. For anything with client data, open the dashboard first (it's at localhost:20128 once it's running) and check which provider it's about to route to. See omniroute.online →
Task Observer · keeps your skills improving
Once you've built a few skills they freeze. They never learn from how you actually use them. Task Observer watches your sessions, notices where you correct Claude or repeat yourself, and writes up suggested changes to your skills. It doesn't touch them itself. You review, you apply. The author calls it "One Skill to Rule Them All", and that phrase triggers it too.
There's no installer for this one. The prompt drops the folder in place and adds one line to your rules file so it actually switches on. At the end of a session, ask it this:
The catch: installing it is not the same as switching it on. If there's no skill-observations folder after a few real sessions, it isn't running, so ask Claude to check the line in your rules file. And if you only have two or three skills, Claude's built-in memory already covers most of this. It earns its keep once you've got a proper stack. See it on GitHub →
Take it all out again
Nothing here is permanent. If you decide you don't want any of it, or you want to start clean, paste this and Claude removes all five and puts your settings back.
Keep the tools that solved a real problem for you and bin the rest. Most people end up with Setup, memory and one of the usage fixes. That's a good place to land.
Ready to go deeper?
Where to go from here.
Wright Mode Membership
Join a community of women entrepreneurs implementing AI and automation in their businesses. Live calls, templates, and ongoing support.
Claude Masterclass
Learn how to use Claude like a pro. From prompting fundamentals to building real workflows that save hours every week.
Claude Code Masterclass
Go beyond the chat interface. Build automations, process data, and create tools with Claude Code — no developer experience needed.