Claude Code Mods: 8 You Should Install Today (Free)

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: Claude Code Mods: 8 You Should Install Today (Free) · Channel: Hyperautomation Labs · 04:39 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Claude Code, Prompt and context engineering, AI safety

Summary

The video introduces eight free Claude Code mods: three from Anthropic and five created by the presenter. The mods help monitor memory use, prevent risky actions, notify users when jobs finish, explain completed work, review file edits, and speed up common prompts. The presenter also demonstrates creating a custom mod by describing it in plain language. Mods have the same computer access as the user, so the video recommends installing only trusted mods and running Claude plugin validate before installation.

Key points

Mods change what Claude Code displays or is allowed to do; skills change what Claude knows.
Claude Code mods require version 2.1.287 or newer and are enabled by default.
Token weather shows memory use and helps identify when to start a fresh chat.
Blast radius previews the files affected by deletions and other destructive commands, allowing the user to proceed or cancel.
Safety net blocks actions such as reading a .env file, erasing a database table, or running a script from the internet.
A completion chime alerts the user when a job lasting more than 30 seconds is done.
Plain English adds a simple sentence explaining work Claude completed.
Replay theater lets users step through file changes; quick buttons and better help with repeated tasks and turning rough ideas into clearer prompts.
Users can create a custom mod by describing it to Claude Code.
Mods have the user’s computer access; install only trusted mods and run Claude plugin validate before installing.

Chapters

Start	Chapter	Summary
00:00	What mods are and what they can do	The presenter explains the problems mods address, how mods differ from skills, and the version requirement.
01:00	Memory monitoring and safety	Token weather tracks memory use, while blast radius and safety net help prevent destructive or risky actions.
02:04	Notifications and clearer answers	The safety net example concludes, followed by mods that notify users when long jobs finish and explain completed work.
03:07	Reviewing changes and improving prompts	Replay theater, quick buttons, and better make it easier to inspect edits, run common requests, and refine rough ideas.
04:08	Making and safely installing mods	The presenter creates a custom mod, explains how to validate mods safely, and describes the free pack.

Quotes

A skill changes what Claude knows. A mod changes Claude code itself. — 00:30

When it says storm, start a fresh chat. — 01:00

Start the job, walk away, and come back when you hear this. — 02:35

you don’t need to write code to make your own. — 04:08

Only install mods from people you trust. — 04:08

Concepts and entities

Claude Code, Anthropic, mod, skill, token weather, blast radius, safety net, plain English, Replay Theater, quick buttons, better, Haiku, Claude Plugin Validate

Full transcript (original language)

00:00 You ask Claude code to do one job. You switch tabs. 20 minutes later, you come back and it finished 19 minutes ago. Or worse, it ran a command you couldn’t read, opened the file with your passwords, or filled up its memory and forgot what you asked for. You end up baby sitting a terminal. Since October 1st, Claude code has mods. A mod is a small add-on that changes what Claude code shows you and what it’s allowed to do. I tested eight. Three are

00:30 Anthropic’s own and five I built for this video. Everyone runs for real on screen. They’re all free and at the end, I’ll show you how to make your own with one sentence. A mod can live in four places. A strip above the box where you type, a panel on the side, a small pop-up note, or a gate that every command passes through before it runs. A skill changes what Claude knows. A mod changes Claude code itself. You need version 2.1.287 or newer and mods are on by default.

01:00 Mod one is Anthropic’s token weather. Claude has a working memory. When it fills up, Claude squeezes old messages down and starts losing details you gave it. This mod puts a weather report for that memory right above your prompt. Here it starts clear, 0%. I ask it to read a big recipe archive and it jumps to cloudy 37%. One more question, 47. The little bars show how much each request added so you can see exactly which question ate your memory. When it says storm, start a fresh chat.

01:31 Mod two, also from Anthropic, blast radius. I asked Claude to clean up my project and it tried to delete a whole folder. Blast radius stopped it and showed me exactly what would be lost. 84 files, about 600 kilobytes listed by name. Then I choose, proceed or cancel. I pressed cancel and Claude confirmed nothing was deleted. It does the same for force pushes and for commands that throw away unsaved work. Blast radius covers deleting. Mod 3 covers the rest. I call it safety net.

02:04 In one message, I asked Claude to read my .env file where my secret keys live, then to erase a database table, then to run a script straight from the internet. Safety net stopped all three. Claude couldn’t see the key and the table was still there afterwards. One thing I learned while testing, Claude code already refuses to wipe your whole project folder on its own. Safety net covers the gaps. Mod 4 ends the babysitting. Dun dun dun. When a job takes longer than 30 seconds, it plays a soft chime and tells you how

02:35 long it took. Here, Claude built a whole order page. 43 seconds, two file edits. Start the job, walk away, and come back when you hear this. Mod 5 is for anyone who reads Claude’s answer and thinks, “Okay, but what did you actually do?” Plain English adds one sentence under every answer where Claude did real work. Written for someone who has never coded. You’ve built a complete ordering system. Open it in a browser and test it. It uses the small, fast Haiku model, so it shows up in about a

03:07 second. Mod 6 is Anthropic’s replay theater. After Claude edits your files, type {slash} replay and step through every change one at a time. Green is what was added, red is what was removed. It’s like rewinding a video of what Claude just did. Mod 7, quick buttons. Type {slash} buttons and you get a panel of the things you ask every single day. Explain it simply. Test it and fix it. Save a checkpoint. What changed? Find my bugs. What’s next? I tap what changed and

03:38 Claude lists every file with one plain line each. No typing. And you can swap in your own six. And mod 8, better. Type {slash} better and a rough idea, like make the order page nicer. It rewrites that into a clear prompt. The goal, the details, how you’ll know it’s done, and a request for a plan first. It lands in your box so you can edit it before you send. Now the best part, you don’t need to write code to make your own. Open Claude code and describe it. Make me a mod that

04:08 shows how long I’ve been working today above the prompt. Claude writes it in under 20 seconds and asks to reload. I say yes and there it is, working today, 0 minutes. The clock just started. Installing mine takes two commands and they’re in the free pack. One warning, a mod runs with the same access to your computer that you have. Only install mods from people you trust. And before you install one, run Claude plugin validate on it. It lists everything the mod touches without running it. Comment mods, all one word, and my bot

04:39 sends you the whole pack. The five mods I built, links to Anthropic’s three, the install commands, and the one sentence prompts to build your own. New to Claude code? My beginner guides are in the description. And if you want to back the channel, become a member here on YouTube and a subscriber on Facebook. If this saved you from babysitting a terminal, hit subscribe.