claude-handoff Skill for Claude Code: Hand Off Work to a Background Agent

Writes a summary of your current Claude Code chat and starts a fresh background agent that picks up the work from it.

Installs: 446.7K; on skills.sh
Stars: 309.2K; on GitHub
Files: 1; besides SKILL.md
Updated: Nov 2026; last repo commit

claude-handoff is a short, user-invoked skill. You type /claude-handoff, optionally with what the next session is for. Claude writes a handoff summary of the conversation, saves it to your OS temp folder, then runs claude --bg --name "<name>" -- "$(cat <file>)" to start a background agent seeded with it. The agent starts in your current folder and the command returns at once. You manage it with claude agents.

The summary includes a "suggested skills" section, skips anything already kept in specs, plans, issues or commits (it links to those instead), and redacts keys, passwords and personal data. The skill sets disable-model-invocation, so Claude will not run it on its own. It sits in the repo's in-progress folder, which the plugin excludes.

Good for

Claude Code users whose session is long and needs a fresh start
Developers who want to run a follow-up task in the background
People who want a clean summary instead of re-explaining context

Install claude-handoff

Claude Code

Copy this prompt.

Install the claude-handoff skill in Claude Code. Run this command:

npx skills@latest add mattpocock/skills --skill=claude-handoff

Then check that it loaded and tell me how to use it.

2
Paste it into Claude Code. It shows the change and asks before running it.

Requires Node.js, Claude Code with the claude --bg command.

The in-progress README says these beta skills are left out of the plugin and the top-level README, so install this one by name. It has no sibling skills. It uses claude --bg, which needs a Claude Code version that supports background agents.

From the project README. Source.

What's in the skill

When it loads

Hand the current conversation off to a fresh background agent that picks up the work immediately.

Key facts

Author: Matt Pocock
Repository: mattpocock/skills, folder skills/in-progress/claude-handoff
Installs on skills.sh: 446,748 as of 2026-11-02
GitHub stars: 309,237 for the whole mattpocock/skills repo, 2026-11-02
License: MIT
Last repo commit: 2026-11-01
What is inside: 1 bundled file besides SKILL.md (agents/openai.yaml); no scripts
skills.sh audits: Gen Agent Trust Hub: pass, Socket: pass, Snyk: pass

What it includes

Starts the handoff; accepts a short focus note
You call it yourself. The argument says what the next session is for.
Try: “/claude-handoff finish the migration tests”
Runs code · /claude-handoff
Saves a handoff summary to the OS temp folder
Skips content already in other artifacts and links to them instead.
Write · Write the summary
Removes API keys, passwords and personal data
The summary becomes the new agent's prompt, so SKILL.md asks for redaction.
Read · Redact secrets
Runs claude --bg with a descriptive name
Passes the summary through a file so the shell does not expand backticks or dollar signs.
Runs code · Launch the agent
Names skills the next agent should call
Added as a section of the summary.
Read · Suggested skills

From the official docs

Permissions

Medium risk

Claude writes a conversation summary to a temp file and starts a background agent in your folder that keeps working. Give it a clear focus and check it with claude agents.

What it can do

Read your data
Create and change things
Delete things
Read-only mode

Watch out for

It starts an agent that keeps working on its own
SKILL.md says the command returns immediately while the agent keeps going. Check it with `claude agents` and give it a clear focus in the argument. github.com
The summary may hold sensitive details
The summary file sits in your temp folder and becomes the new prompt. SKILL.md asks for redaction, but check it if the chat held secrets. github.com

Security advisories

No published security advisories as of Nov 2, 2026.

Technical details

Turning on read-only	No read-only mode. The skill only starts when you type /claude-handoff.
Talks to	The new agent sends its work to your AI provider like any other Claude session
Sign-in and keys	SKILL.md mentions no key or account; the new agent is started with the claude command.
Scopes	Writes a summary file to your OS temp folder Runs the claude command in your shell to start a background agent The new agent runs in your current working folder

Questions

How do I install claude-handoff?

Use the install box at the top of this page. The repo README says in-progress skills are not in the plugin, so you install this one directly with the skills command.

What does the new agent get?

It gets the handoff summary as its prompt, starts in the same working folder, and has a name you can see in claude agents.

Will Claude run it automatically?

No. The skill sets disable-model-invocation: true, so only you can start it by typing /claude-handoff.

How is it different from the handoff skill?

The README describes handoff as compacting a conversation into a document another agent can continue from. claude-handoff goes one step further and launches that agent for you with claude --bg.

Is it safe?

Use care. It writes a file to your temp folder and starts a background agent that can act in your folder. SKILL.md asks Claude to redact secrets, but read the summary if the chat held sensitive material.

Information

updated Nov 2, 2026

Repository: mattpocock/skills
Folder: skills/in-progress/claude-handoff

mattpocock/skills skills.sh

Sources

Docs, issue trackers and discussions this page is built from.

Show all 3

in-progress README · github.com
claude-handoff SKILL.md · github.com
mattpocock/skills README · github.com