cc-plugin-you-should-know: what this Claude Code mod does

cc-plugin-you-should-know is a mod built into Claude Code that keeps a second agent watching while Claude works on longer tasks. When that agent spots something you or Claude might miss, it shows you a short note above the prompt. Its display name is You should know; cc-plugin-you-should-know is what it is called in the /plugin list, which is why that is the name people search. It is off by default. You turn it on with /plugin enable cc-plugin-you-should-know@builtin.

Why it has such an odd name

The feature is called You should know. It shipped in Claude Code v2.1.287 on October 1, 2026, the same release that added mods.² The name you actually see, and probably searched for, comes from how Claude Code lists its own built-in mods.

Open /plugin and go to the Installed tab. Built-in mods are listed under Built-in, each with a cc-plugin- prefix: cc-plugin-diff, cc-plugin-agents-md, and cc-plugin-you-should-know.¹ The @builtin part of the enable command says where the plugin comes from. For a plugin you install it would be the name of its marketplace. For these built-in ones, it means the plugin ships inside Claude Code.⁴

So cc-plugin-you-should-know@builtin just means: the You should know plugin, the one that comes with Claude Code.

What it does, and when a note appears

Anthropic's docs describe it in two parts.¹ First, it runs a side agent: a second agent that works next to Claude, not instead of it. That agent watches while Claude works on longer tasks. Second, when it finds something worth knowing that you might miss, it shows you a note above the prompt. The release notes put the goal as flagging things that you or Claude might miss.²

Think of a colleague reading over your shoulder during a long task. Most of the time they stay quiet. Now and then they lean in and say: you should know this.

One detail came a day later. In v2.1.288, the notes started to say "we", "the main agent" or "you", depending on who was responsible for a decision.³ So a note can say who made a decision: Claude, or you.

What Anthropic has not published is the exact rule for when a note shows up. The docs do not say how long a task must be to count as longer, which model the side agent uses, or how often it checks. The other built-in mods have their source in the public mods folder of the Claude Code repository. This one does not.⁵ So treat it as a helper whose timing you will learn by using it.

How to turn it on and off

You need Claude Code v2.1.287 or later. Run claude --version to check. Then, inside a session, type:²

> /plugin enable cc-plugin-you-should-know@builtin

That command opens the Installed tab of /plugin at the plugin and turns it on.⁴ You can also do it by hand: open /plugin, go to Installed, choose Show disabled, and enable it there.¹

If you cannot find it, there are a few likely reasons. It is only listed if it is available for your organization.¹ The release notes say it is for first-party sessions with telemetry on.² And /plugin only works in an interactive terminal session, not in claude -p.⁴ Or you are on a version older than v2.1.287.

To turn it off, disable it in /plugin, on the Installed tab.¹ The command form works the same way:⁴

> /plugin disable cc-plugin-you-should-know@builtin

One thing to know: --safe-mode, --bare and the disableAllHooks setting stop the mods you installed, but they do not stop built-in mods.¹ If you want You should know off, disable it in /plugin. You cannot uninstall it, because built-in mods cannot be updated or uninstalled, only turned off.¹

What it costs

There is no separate price. But a side agent is a model, and a model call is not free. Anthropic's mods docs say plainly that a mod can spend your usage: it can call a model on your plan or your API key.¹ So with You should know on, expect some extra usage from the side agent on top of Claude's own work.

Anthropic has not published how much. If you are on a plan with tight limits, or you pay per token on an API key, turn it on for a few longer tasks and compare your usage with and without it before you leave it on.

What it can and cannot see

A mod runs inside your Claude Code session. Anthropic's docs list what any mod can reach: every prompt you send and every tool call Claude makes, plus your files and environment, because mods are not sandboxed.¹ You should know is a mod, and its job is to watch the session, so assume it sees what Claude sees in that session.

What it cannot see is anything outside the session. It does not know what your app is doing in production, which requests are failing for your users right now, or what happened last night. It can only flag what shows up in the conversation and in the tools Claude runs.

Is it worth turning on?

It depends on how you work with Claude Code.

• Probably yes if you give Claude long tasks and step away, or skim the transcript. That is exactly when things slip past you: a choice Claude made on its own, an assumption, a step it skipped. A note above the prompt is easy to catch.
• Probably not if your tasks are short and you read every step, or if every token on your plan counts. The side agent adds usage, and on short tasks there is little for it to catch.

It is easy to try. Turn it on for a day of real work, see whether the notes tell you anything you would have missed, and turn it off in /plugin if they do not.

How it fits with mods in general

A mod is a plugin with a few functions that run inside Claude Code and get a say each time something happens: a prompt, a tool call, the screen being drawn. Anthropic now builds some of Claude Code itself this way. /diff is a mod, AGENTS.md support is a mod, and You should know is a mod.¹

That also means you can build your own. A mod can watch tool calls, draw a band above the prompt, or stop an action and tell Claude why. If mods are new to you, start with our plain guide: Claude Code mods, explained simply, with a first mod. For mods that keep Claude inside rules you set, read rules Claude Code cannot ignore.

You should know watches the session. Nothing in the session watches production. Fixter is monitoring for teams that build with coding agents: you send your logs and traces over standard OpenTelemetry, and Fixter finds the issues and bugs in your system and tells you what broke and why. You pull them into Claude Code over MCP and fix them there.

Frequently asked questions

What is cc-plugin-you-should-know?

It is the name Claude Code's /plugin panel shows for You should know, a mod built into Claude Code since v2.1.287 (October 1, 2026). When you turn it on, a side agent watches while Claude works on longer tasks and shows you a short note above the prompt when it finds something worth knowing that you might miss.

Is cc-plugin-you-should-know safe?

It is made by Anthropic and ships inside Claude Code itself, so you are not installing code from a stranger. Like any mod it runs inside your session and can see what happens there, and its side agent makes model calls that count toward your usage. Unlike the diff and AGENTS.md mods, its source is not in the public mods folder of the Claude Code repository, so you cannot read its code.

How do I enable cc-plugin-you-should-know?

Make sure you are on Claude Code v2.1.287 or later, then type /plugin enable cc-plugin-you-should-know@builtin at the prompt. You can also open /plugin, go to the Installed tab, choose Show disabled, and enable it there. If it is not listed, it is not available for your organization or setup.

How do I disable cc-plugin-you-should-know?

Open /plugin, go to the Installed tab, find cc-plugin-you-should-know under Built-in, and disable it. The command form is /plugin disable cc-plugin-you-should-know@builtin. Note that --safe-mode, --bare and the disable all hooks setting do not stop built-in mods, so disabling it in /plugin is the way.

Why can't I find cc-plugin-you-should-know in /plugin?

It is off by default, so it only shows after you choose Show disabled on the Installed tab, and only if it is available for your organization. The release notes also say it is for first-party sessions with telemetry on. You also need Claude Code v2.1.287 or later; run claude --version to check.