You Should Know in Claude Code: A Feature That Watches for What You Might Have Missed

A long session with an agent almost always ends the same way: somewhere in the middle it made a decision it didn't state directly, or you asked for something yourself and then forgot whether it was done. It usually comes up later, when the job has already been done wrong.

On October 1, 2026, Claude Code version 2.1.287 came out, and with it a built-in plugin, You Should Know, made for exactly this.

What it does

You Should Know starts a separate helper agent that watches the main session and notices what you or Claude itself may miss. When it finds something important, a notification appears right above the line for the next request, so you don't have to scroll through the chat history again.

Anthropic presents it as an agent that keeps watch over you during long tasks. By the wording of the release, the plugin highlights decisions, changes, and warnings, not only technical errors.

How to turn it on

The plugin is off by default. You turn it on with one command right in a Claude Code session:

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

If the plugin was already turned on and then off, you can find it through /plugin → Installed → Show disabled.

Who can use it right now

Honesty matters here: the plugin's availability is narrow. It works only in first-party Claude Code sessions with telemetry on. If telemetry is off or the session isn't going through the official Anthropic client, the plugin is unavailable, regardless of version.

A day after the release, on October 2, 2026, version 2.1.288 came out with a clarification of the notification wording. The mechanism itself didn't change. What was fixed is how the plugin names who made a decision: "we", "the main agent", or "you".

Is it worth turning on

The upside is clear: in long agent sessions it's easy to lose the thread, especially if the task stretches over several hours and outside edits got mixed in. A separate watching agent lowers the cost of such a loss.

The downside is clear too and directly tied to the access condition: you can turn on the plugin only if you consent to telemetry. This is not development in a closed loop with no data passed to Anthropic, so the choice should be made consciously, not by turning on a feature just because it's new.

In my content factory, Claude Code sessions often stretch over several hours, with edits from different people in one stream, and that is exactly where a decision gets lost first. I keep telemetry on in the working project anyway, so I'll turn this plugin on and watch what it highlights over the first week before deciding whether to keep it permanently.

Bottom line

You Should Know closes a specific gap: an agent can do something important and not highlight it explicitly, and the person finds out only after the fact. The plugin adds a watching layer right into the interface, but for now it isn't available to everyone and requires telemetry to be on. For those who already work in the official client with telemetry, turning it on with one command won't hurt.

I write about working practices in Claude Code in the channel «Растём с ИИ» ("Growing with AI"). A breakdown of a related topic: how to write a skill for Claude Code so it really works.

If you need to go through your own system of agents in person, you can start with a free consultation.