← All projectsSKILL / OPEN SOURCE

claude-next

Your next request.
In its own turn.

Leave follow-up requests with /next while Claude works. Each runs in its own turn after the current task ends.

01 / CLAUDE CODE
Claude Code
# Leave the next request while work continues.

❯ /next run the tests

↳Runs in the next turn when this task finishes
/next → queue → next turn
LUCIANLABS / CLAUDE-NEXTBuilt to be used. And shared.
HOW IT WORKS

How to use it.

01

Leave a request.

While Claude Code works, type /next followed by your next request.

02

Finish the current turn.

The request waits for the current turn to end. It does not wait for background tasks.

03

Continue in a new turn.

Queued requests run one per turn, in order. Open a new Claude Code session after installation.

IN DEPTH

In depth

Why /next exists, how it waits, and what it does not do.

Why it exists

When a follow-up comes to mind while Claude is working, most people just type it. But a normal message sent mid-task reaches Claude as soon as the current tool call finishes. Send “add a test for an empty list” while it is fixing a bug, and it lands before the tests even run, so both requests end up in one reply.

Slash commands, on the other hand, wait until the turn ends, as Claude Code’s documentation describes. /next is a one-line skill built on that difference.

How it works

The whole skill is a few lines. It hands whatever follows /next to Claude as the next request in the conversation; the waiting is done by Claude Code’s own command queue.

---
name: next
description: Queue a follow-up request that runs after the current turn finishes.
argument-hint: <next request>
disable-model-invocation: true
---

Treat the following as the user's follow-up request and carry it out now,
continuing from the current conversation: $ARGUMENTS

Because of disable-model-invocation: true, only you can run it. It never appears in Claude’s skill list, so it costs no context until you use it.

Using it

  • While Claude works, type something like /next run the tests and fix anything that fails. It runs as a new turn once the current one ends.
  • To queue several, press Enter after each one. They run in order, one turn each.
  • To cancel, press ↑ with the input empty, delete the line and press Enter. Don’t use Ctrl+C for this; it stops Claude.

How it was tested

The repository’s tests/verify.sh drives a real Claude Code session in tmux, queues two /next commands during a multi-step task, and reads the session transcript to see what reached Claude and when. On Claude Code 2.1.283 (macOS), neither request reached Claude during the first turn (0 leaks), and each ran in order as its own turn.

Limits

  • It waits for the main turn to end, not for background tasks Claude started.
  • Permission prompts still apply.
  • It is a personal skill, so it works in local sessions on your machine, not in cloud sessions.

FAQ

How do I update it?

Run npx claude-next-skill again.

How do I uninstall it?

Run npx claude-next-skill uninstall.

Can I use it without Node.js?

Save skills/next/SKILL.md unchanged as ~/.claude/skills/next/SKILL.md (on Windows, %USERPROFILE%\.claude\skills\next\SKILL.md).

GET STARTED

Try it in
your workflow.

Install in your terminal, then open a new Claude Code session.

TERMINAL
npx claude-next-skill
Read the full guide on GitHub ↗
← Explore other projects