These 9 Claude Mods Make Claude Code Feel Brand New (Install Guide Included)

Author: Hyperautomation Labs · Watch on YouTube

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: These 9 Claude Mods Make Claude Code Feel Brand New (Install Guide Included) · Channel: Hyperautomation Labs · 08:23 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Cybersecurity and software security, Claude Code, Prompt and context engineering

Summary

The video introduces nine Claude Code mods that change what Claude can do and what users see, including tools for testing, safety, usage tracking, memory, file visibility, secret protection, and notifications. It demonstrates how each mod works, notes costs or limitations, and explains where Claude Code mods run. The presenter also covers building a custom mod, troubleshooting common issues, and validating a mod before installing it. Install lines are said to be included in the video description.

Key points

Mods change Claude Code itself, while skills teach it to do a job and connectors link it to other apps.
Mods require Claude Code version 2.1.287 or newer and work in the terminal, the desktop app’s code tab, and VS Code.
Proof of Done runs tests after Claude changes files and can send Claude back to fix failures.
Blast Radius previews the effects of certain commands, but it is not a complete safeguard and does not monitor file edits.
Cache Tags warns about the cost of a cold conversation reread, while Keep Warm can send a small ping before the memory goes cold.
Usage Meter displays conversation memory and usage limits using numbers Claude Code already has.
Reflect can save corrections as project instructions, and a separate memory mod can recap unfinished work when a session resumes.
File Tree shows file activity, while Secret Shield replaces keys with placeholders before they leave the computer.
Phone Buzz sends notifications when Claude needs input or finishes a long job.
The video recommends using Claude Plugin Validate before installing a mod and gives basic troubleshooting steps.

Chapters

Start	Chapter	Summary
00:00	What mods are and where they work	The presenter defines Claude Code mods, explains the version and platform requirements, and warns that mods run with access to your computer.
01:33	Testing and command safety	Proof of Done catches failing tests after changes, and Blast Radius previews certain command effects while remaining an imperfect safety net.
02:37	Usage tracking and conversation memory	Cache Tags and Keep Warm address cold conversation memory, while Usage Meter displays session and usage-limit information.
04:11	Memory, file visibility, and secret protection	The presenter demonstrates mods that save corrections and session context, show file activity, and replace secrets with placeholders.
06:18	Notifications, custom mods, and troubleshooting	Phone Buzz sends alerts, the presenter explains how to build and install a custom mod, and the video closes with troubleshooting and plugin validation advice.

Quotes

A skill teaches Claude how to do a job. A connector plugs Claude into another app. A mod changes Claude code itself — 00:31

The mod itself costs nothing to run. It’s plain code, not another AI. — 01:33

It’s a safety net, not a lock. — 02:06

Phone buzz pings your phone the moment Claude needs a yes or asks you a question. — 06:18

Before you install anything, run Claude Plugin Validate on it. — 07:20

Concepts and entities

Claude Code, Proof of Done, blast radius, Cache Tags, Usage Meter, Reflect, File Tree, Secret Shield, Phone buzz, Claude Plugin Validate

Full transcript (original language)

00:00 Claude just told you it’s done. It isn’t. Your tests are failing. You went to make a coffee and Claude has been sitting there the whole time waiting for you to press yes. And the next message you send is going to eat a big chunk of your usage limit. Because while you were away Claude’s short-term memory of your conversation went cold. Every one of those has a fix now and they’re called mods. This is Claude code with no mods. And this is the same

00:31 Claude code with nine. By the end of this video you’ll know exactly which nine to install. What each one really looks like on your screen, what it costs you and where it works. I installed every one of them for this video and every install line is in the description with no sign-up. 30 seconds on what a mod is. A skill teaches Claude how to do a job. A connector plugs Claude into another app. A mod changes Claude code itself, what

01:01 you see on the screen and what Claude is allowed to do. Mods are switched on by default. You just need Claude code version 2.1.287 or newer and if you’re not sure, type Claude update. They work in the terminal in the code tab of the desktop app and now in VS code too, which wasn’t true when mods first came out. One rule before we start. A mod runs with your access to your computer. So only install mods from people you trust. I’ll show you how to check any mod in 10 seconds at the end.

01:33 Let’s fix that first problem. Proof of done runs your tests every time Claude changes your files. Here, I asked Claude to rename one function and touch nothing else. It said it was finished and the mod caught a failing test. Now Claude doesn’t get to stop. The mod sends it straight back with the exact failure and I don’t type a word. It fixed its own code, it left my test alone and now seven out of seven pass. The mod itself costs nothing to run. It’s plain code, not another AI.

02:06 Next, the one that might save your whole project. Blast Radius was made by Anthropic’s own team as a sample. I asked Claude to delete my old build folder. Before anything ran, the mod stopped it and showed me exactly what would disappear. 84 files. Proceed or cancel? I pressed cancel, and Claude told me nothing was deleted. Its own instructions are honest, and I will be too. It’s a safety net, not a lock. It doesn’t catch every possible way to delete files, and it only watches commands, not file edits. So, keep a

02:37 real backup of your project. This is the seatbelt on top. Now, the ones that save your usage. While you chat, Claude keeps a short-term memory of your conversation, so it doesn’t have to reread all of it every time. Step away for over an hour, and that memory goes cold. Your next message makes Claude reread everything, and that’s what eats your limit. Cache Tags stops that first cold message once and shows you what it’s about to cost before you spend it.

03:10 Here, I came back to a long session after more than an hour. One quick recap question would have reread 350,000 tokens, about $1.40 at API prices, against 4 cents for a warm turn. Going to lunch? Type keep warm and a time, and it sends a tiny ping just before the memory goes cold. Those pings use a little usage, but much less than one cold reread. The question I asked myself 10 times a day was simple. Am I about to hit my

03:40 limit? Usage Meter puts the answer right under your prompt in every session. How full Claude’s memory of this conversation is. How much of your 5-hour limit you’ve used, and when it resets. and how much of your week is gone. The It only reads numbers. Claude Code already has, so it costs you nothing. I built this one, and it’s in the free pack at the end. If this is already saving you time, hit subscribe because it’s free. And if you want to back the work, hit join here or

04:11 the supporter button on Facebook. Now, the memory ones. You correct Claude, it says sorry, and next week it does the exact same thing. Reflect catches the moment you correct it and asks if you want to keep that as a permanent rule. Here, I told Claude that prices are always in cents in this project. The mod spotted the rule. I pressed save, and it went straight into Claude’s instructions file for this project. So, next time it already knows. It uses a small cheap model to spot your corrections. So, it

04:42 does cost a little. And it works the same if you write emails instead of code. Tell it never to use exclamation marks once, and that’s the last time. Someone in the comments asked for a mod that just remembers basic facts about the project. This is that. Close your laptop tonight, open it tomorrow, and Claude greets you with one line: what you were doing, what’s finished, and what’s still open. Press one button, and it picks up right there. Here, it even noticed the muffin I added last time had no test, and it

05:13 wrote one. Eight out of eight pass. File tree gives Claude Code a file panel. Files glow purple while Claude reads them, orange while it writes, and green when the change is committed. So, you see what it touched at a glance. If you’re not a developer, picture a folder of tax documents, and you can see exactly which ones Claude opened and which ones it changed. And this answers the number one complaint I saw. Mine doesn’t look like the video. In the terminal, this panel only shows in full screen mode in a wide enough

05:46 window. Type {slash} tui full screen and make the window wider. In the desktop app, it docks on its own. Claude reads your files, and that includes the ones with your passwords and API keys. Secret Shield swaps every key for a placeholder before anything leaves your computer. I asked Claude to read my payment key back to me. It can’t. It never saw it. Think of it as a seatbelt, not a safe. So, still keep real keys out of your code.

06:18 End the coffee problem from the start. Phone buzz pings your phone the moment Claude needs a yes or asks you a question. And again, when a long job is done. It uses a free notification app. No account and no sign up. You just subscribe to your own private channel once. You don’t need anyone’s report to get a mod. Describe what you want in one sentence. I asked for a card that shows how long Claude took and how many files it changed. Claude code has a built-in skill for building mods, so it knows

06:49 exactly how. When it asks about hot reloading, say yes and the mod shows up in the same session. Don’t like it? Tell it what to change the way you tell a person. One catch. A mod Claude builds is temporary. When you love one, ask Claude to copy it out and install it as a plugin and it’ll be there every time you open Claude code. If a mod won’t work, it’s almost always one of four things. If you see an error about an unrecognized key called options, your Claude code is too old, so

07:20 run Claude update. If it installed but nothing shows, type {slash} reload plugins or restart. If it looks different from the video, switch the terminal to full screen and widen the window. And if a mod forgets its settings when you type clear, that’s how clear works. So, check that mods page for a fix. And the 10-second check I promised. Before you install anything, run Claude Plugin Validate on it. It lists everything the mod can touch without running it. Real talk, 90% of you watching haven’t subscribed. On YouTube, hit join and

07:53 become a member. On Facebook, subscribe is the supporter button, and that’s what keeps the lights on here. And if a paid membership isn’t for you, a free YouTube subscribe still helps. Subscribe to support us. Every install line is in the description right now. If you want the whole pack with my five mods, the exact sentences that built them, and the fixes for when something breaks, comment modded and I’ll send it to you. New to Cloud Code? My beginner guides are in the description. And tell me in the

08:23 comments, which one are you installing first? Follow along on Instagram and Facebook for bite-size AI, and the deep dive playbooks live here on YouTube. So, subscribe so you don’t miss one. I’ll see you in the next one.