File of the day
A Claude Code rule that marks every task and problem by urgency
Slava Sazhin 5 October 2026 4 min read 0 views
A rule for your global CLAUDE.md: every task and problem Claude lists for you starts with a colored mark, so you see at once what to do first.
Today's file is a rule for Claude Code. Ask Claude to check a plan or a script and you get a list of tasks and problems in one even tone: a deadline that runs out tomorrow looks the same as a nice-to-have idea. With this rule every item starts with a colored mark, so you see at once what to do first.
How it works
There are five marks. 🔴: losses are happening now or will start before tomorrow, such as work has stopped, orders do not arrive, money is leaking. 🟠: losses are already happening, but slowly, and it can wait until tomorrow. 🟡: no losses yet, but they will come: a deadline ahead, a task nobody watches, a supply running out. 🟢: an idea or an improvement, whatever its size. ✅: good news, and only when you ask to hear it. The three urgent colors are set by the cost of delay, not by how interesting or hard the task is, and a task that waits for your decision is rated like a problem. Claude puts one mark per item and does not repeat the urgency in words.
In our test we asked in four languages to check a plan for a coffee shop: the rent is paid until the end of the month, the coffee machine arrives in a week, there is no cash register yet, the sign gets ordered later. Without the rule there were no marks. With it every item had one: the rent, the machine and the sign got 🟡, the missing cash register got 🟡, and in some runs 🟠, because the rent is already being paid while the shop cannot sell. Ideas such as a temporary banner until the sign is ready came as 🟢, and Claude's own question "what is your opening date?" got a mark too. To "what is 17*23" Claude answered 391 with no mark. If you also have yesterday's premortem rule, the two work together: the "What could go wrong" items keep their numbers and get a mark after the number.
What to change
The file is a heading, the list of five marks and a few short lines on how to put them. Change the examples in the list to fit your work: "orders do not arrive" suits a shop, a freelancer may want "a client is waiting". To see good news with ✅ every time, remove "Put only when I ask to hear about the good". We added one line ourselves. In the test the rule also put marks into a todo.md file Claude wrote and into the strings of a JSON answer, though we asked for JSON and nothing else. A script that reads such output, for example from a claude -p run on a schedule, would get "🟡 The rent..." instead of "The rent...". Now the file says marks go only into answers in the chat: files, code, commit messages and output that a program reads (JSON, CSV) get none, even when that output is the answer. After that 12 JSON answers out of 12 in four languages came back without marks, and todo.md had none, while the chat answer about the same list kept them.
Install
The rule goes at the end of ~/.claude/CLAUDE.md, the file with your instructions for all projects. If you do not have it yet, Claude creates it; the note below has the exact request. That folder holds Claude's own settings, so Claude asks before it edits the file: answer Yes. Then close Claude Code and open it again, because the file is read at startup. The check from the note below: ask Claude to check the coffee shop plan, and every task and every problem in the answer starts with a colored mark.
The files
Files of the day are for paying subscribers: any plan opens every one. Already subscribed? Then sign in.
See the plansWas this article worth your time?
Read next
File of the day
A Claude Code rule that keeps a task log so a new conversation continues where you stopped
A rule for your global CLAUDE.md: Claude keeps a log of each task, and a new conversation reads it and continues where the last one stopped.
Claude Code cloud sessions or your own server: which one for which job
Cloud sessions come with your Claude plan and suit a coding task on GitHub. An agent that has to stay on all day needs a machine of its own.
What I mean by an AI-native company
My answer to what an AI-native company is: one where colleagues are not split by which AI tools they have and how well they use them.
Give Claude a computer of its own
Your own Linux machine with Claude Code on it, working around the clock while your laptop is shut. From $3.49 a month.