claude-handoff
vinvcn/mattpocock-skills-zh-cn
Hand off the current conversation to a new background agent for continued work.
What is claude-handoff?
Creates a handoff summary of the current conversation and launches a new background agent to continue the work. The skill generates a concise summary with suggested skills for the next agent, then starts it as a background process that the user can manage with `claude agents`.
- Generates a handoff summary capturing context and progress from the current conversation
- Launches a new background agent with a descriptive name using the summary as its prompt
- Includes a 'suggested skills' section recommending which skills the next agent should use
- Removes sensitive information (API keys, passwords, PII) from the handoff summary
- Accepts optional arguments to focus the next session on specific priorities
- Avoids duplicating content already in artifacts (PRDs, plans, issues, commits) by referencing them instead
How to install claude-handoff
npx skills add https://github.com/vinvcn/mattpocock-skills-zh-cn --skill claude-handoffHow to use claude-handoff
- 1.Invoke the skill when you want to hand off work to a background agent
- 2.Optionally provide arguments describing what the next session should focus on
- 3.The skill generates a summary of the current conversation and context
- 4.A new background agent launches automatically with a descriptive name
- 5.Use `claude agents` command to monitor and manage the background agent
Use cases
- Delegate a long-running task to a background agent while you continue with other work
- Hand off a partially-completed project to another agent with full context
- Split complex work across multiple specialized agents working in parallel
- Continue work on a task without interrupting your current conversation flow
- Manage multiple concurrent agent tasks with descriptive names and status tracking
- Developers managing multiple concurrent coding tasks
- Teams coordinating work across multiple agents
- Anyone needing to parallelize long-running background work
- Users working with Claude Code or Cursor agent environments
claude-handoff FAQ
The summary includes conversation context, progress made, and suggested skills for the next agent. It excludes sensitive data like API keys and passwords, and references existing artifacts (PRDs, plans, commits) by path rather than duplicating them.
Yes, you can pass arguments to the skill describing the focus for the next session, which will be incorporated into the handoff summary.
Use the `claude agents` command to view, manage, and interact with background agents that are running.
The background agent receives the handoff summary as its prompt, which captures the essential context and progress. It will not have access to the full conversation history unless explicitly referenced in the summary.
No, the skill automatically removes sensitive information like API keys, passwords, and PII before creating the background agent, as the summary becomes the agent's prompt.
Full instructions (SKILL.md)
Source of truth, from vinvcn/mattpocock-skills-zh-cn.
name: claude-handoff description: 把当前对话交接给一个全新的 background agent，让它立即接手工作。 argument-hint: "下一个 session 将用于什么？" disable-model-invocation: true
为当前对话写一份 handoff summary，让一个全新的 agent 可以继续工作。不要把它保存下来，而是启动一个以这份 summary 作为 prompt 的 background agent：claude --bg --name "<descriptive name>" "<handoff summary>"。它会在当前工作目录中启动并立即返回；用户用 claude agents 管理它。
始终带上 -n/--name 并给出描述性名称（例如 --name "Fix login bug"）——它设置显示名称，会出现在 job list、session picker 和 terminal title 中。
在 summary 中包含一个 "suggested skills" section，写明下一个 agent 应调用 Skill 工具并指定哪些 skills。
不要重复其它 artifacts（PRDs、plans、ADRs、issues、commits、diffs）中已经记录的内容，改为用 path 或 URL 引用它们。
删去任何敏感信息，例如 API keys、passwords 或个人身份信息——这份 summary 会成为 agent 的 prompt。
如果用户传入了 arguments，把它们当作对下一个 session 重点的描述，并据此调整 summary。
Related skills
More from vinvcn/mattpocock-skills-zh-cn and the wider catalog.
code-review
Dual-axis code review: check changes against repo standards and spec requirements in parallel.
codebase-design
Shared vocabulary for designing deep modules with small interfaces and hidden complexity.
design-an-interface
Generate multiple radically different interface designs for a module using parallel sub-agents.
diagnose
Disciplined debugging loop: reproduce → minimize → hypothesize → instrument → fix → regression-test for tricky bugs and performance regressions.
diagnosing-bugs
Structured diagnostic loop for tricky bugs and performance regressions.
domain-modeling
Build and refine your project's domain model through active terminology and decision capture.