Turn Any Repeating Claude Prompt Into a Slash Command

Rewriting the same context block every Monday morning adds up faster than you’d think. Same priorities, same stakeholders, same framing for the week. You’re essentially introducing yourself to a tool that could already know all of this. Claude Code skills fix that, and building one takes about ten minutes.

Your new slash command

You’re building /weekly-brief: a personal Claude Code skill that takes a quick brain dump about the week and returns a focused brief you can reference for the rest of that session. The skill stores your instructions once, and you invoke it on demand.

Claude Code skills are different from saving a prompt in a note somewhere. A skill loads only when called, lives outside your conversation history, and is available in every project you open. Write it once and it’s just there.

Building it

Skills live in your home directory under ~/.claude/skills/. Each skill is a folder with a SKILL.md file inside: Claude Code reads the folder name as the command and the file as the instructions.

Create the folder:

mkdir -p ~/.claude/skills/weekly-brief

Save this to ~/.claude/skills/weekly-brief/SKILL.md:

---
description: Generate a weekly context brief from bullet points. Use when the user shares weekly priorities, goals, or what is in motion for the week.
disable-model-invocation: true
---

You are helping me set context for the week ahead.

I will give you a loose brain dump of what is happening. Turn it into:

1. A single sentence summarizing the week
2. Three priorities ranked by urgency
3. One decision I need to make before Friday
4. Any risks worth watching

My input: $ARGUMENTS

Keep it tight. I will reference this context all week.

Three things in that frontmatter are doing real work:

description tells Claude when to load the skill automatically. Write it as a situation, not a label. “Use when the user shares weekly priorities” triggers automatic loading when relevant. “Weekly brief generator” does not.

disable-model-invocation: true means Claude cannot run this on its own. Without it, Claude might generate a brief any time you mention your week. For a skill with a specific output format, you want to decide when it fires.

$ARGUMENTS gets replaced by whatever you type after /weekly-brief. Everything after the command name lands there.

Claude Code watches skill directories for changes. Edit SKILL.md while a session is open and the update takes effect in the same session, no restart needed.

Running it

Open any project in Claude Code and type:

/weekly-brief Q2 close is this week, pricing deck due Thursday, two stakeholder reviews, new team member starts Monday, product update running behind schedule

Here is the kind of brief Claude produces from that input:

---

Week in one line: High stakes close week with a firm Thursday deadline running alongside a new hire start and a flagged delay.

Three priorities:

1. Pricing deck finalized and reviewed before Thursday’s stakeholder session

2. New team member onboarding (starts Monday, protect that first morning)

3. Product update: surface the delay now before it becomes a fire next week

Decision before Friday: Does the pricing deck go to both stakeholders before or after Thursday’s review? Decide Tuesday morning.

Risk to watch: Stakeholder reviews and onboarding land in the same 48 hours. Block Tuesday afternoon now.

---

The brief stays in context for the rest of that session. Ask “what were my three priorities?” three days later and Claude still has them.

Where this breaks

• Vague input produces vague output. “Busy week” gives Claude nothing to rank. The sharper your brain dump, the sharper the brief.

• The brief lives in one session. Close Claude Code and reopen it, and the context is gone. Paste the output somewhere permanent if you want it to survive.

• Personal skills live in ~/.claude/skills/ and don’t sync across machines. Copy the file if you use Claude Code on two computers.

One extension worth adding

Add !date as the first line of the skill body. Claude Code executes that before the skill loads, injecting today’s actual date into the prompt. The brief can then reference real days (“block Tuesday the 2nd”) instead of relative terms like “by Friday.”

I write about building with AI on LinkedIn. Come say hi.