Agent Native

handoff

This skill hands off ongoing work to a fresh agent, either in the same workspace or a new isolated workspace, while preserving context through a synthesized…

by ImbueOfficial publisherNot reviewed

Identity proven by GitHub's verified organisation badge (imbue-ai)

Agents
Claude Code
License
MIT (permissive)
Pin
f847102pinned Oct 9, 2026
Context cost
about 1,100 tokenswhen the skill loads
GitHub stars
236on imbue-ai/sculptor
Scripts
None found

What it does

“Hand off the current work to a fresh agent — either a new agent in the SAME workspace, or a new workspace — seeded with a summary of the current context.”
The author's description, from SKILL.md

Generated summary from the skill's own text

This skill hands off ongoing work to a fresh agent, either in the same workspace or a new isolated workspace, while preserving context through a synthesized prompt. It manages the handoff decision, commits any uncommitted changes if needed, and creates the new agent with relevant task summary and next steps.

When to use it

  • User wants to continue work with a fresh agent context window
  • Large conversation history makes a clean restart helpful
  • Switching to a new isolated workspace for continued work
  • User requests context transfer to resume after interruption

When not to

  • Work should stay with the current agent without interruption
  • No need to preserve conversation context for next steps
  • New workspace is unnecessary for the task at hand

Install

These commands install from the publisher's repository and need no account.

skills CLI

npx skills add imbue-ai/sculptor --skill handoff

Installs from the repository's default branch, which may have changed since the commit we checked.

GitHub CLI 2.90 or later

gh skill install imbue-ai/sculptor handoff@f847102ada1d6896bf6a3a5eccde996760526d9e

Installs the commit we checked (f847102).

Copy by hand

Download the folder at f847102 and put it in .claude/skills/handoff for Claude Code and in .agents/skills/handoff for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode.

Skill folder on GitHub at f847102

Copy for your agent

A prompt for any coding agent. It installs this skill from GitHub at the pinned commit, copies the files as they are, and runs nothing unless you ask.

Prompt

Install the agent skill "handoff" by Imbue.

Source, pinned to commit f847102ada1d6896bf6a3a5eccde996760526d9e:
https://github.com/imbue-ai/sculptor/tree/f847102ada1d6896bf6a3a5eccde996760526d9e/sculptor/sculptor-experimental/skills/handoff

1. Download that folder at exactly this commit. Do not use the default branch.
2. Copy every file unchanged, keeping the folder structure, into .claude/skills/handoff/ for Claude Code and .agents/skills/handoff/ for Codex, Cursor, Gemini CLI, GitHub Copilot and OpenCode. Create the folders if they do not exist.
3. Do not edit the files, and do not run any script in the folder unless I ask you to.
4. When you are done, list the files you copied and where you put them.

Listing and checks: https://www.agentnative.dev/skills/imbue/handoff

SKILL.md at the pin

---
name: handoff
description: |
  Hand off the current work to a fresh agent — either a new agent in the SAME
  workspace, or a new workspace — seeded with a summary of the current context.
when_to_use: |
  Invoke when the user wants to continue the current work with a fresh agent —
  e.g. "hand this off", "start fresh but keep the context", or when the current
  context window is large and a clean restart would help. Works in two modes: a
  new agent in the same workspace (shares the branch and files), or a brand-new
  workspace on its own branch off the current one.
user_invocable: true
---

# Handoff

ARGUMENTS: $ARGUMENTS

Hand off the current work to a fresh agent, seeded with a summary of the current
context. `$ARGUMENTS` (if provided) describes what the new agent should focus on
next; if empty, infer it from the recent conversation.

This skill uses the `sculpt` CLI. The agent shell already has
`SCULPT_WORKSPACE_ID` (current workspace) and `SCULPT_PROJECT_ID` set.

## Step 1 — Choose the destination

The handoff can land in one of two places:

- **New agent, same workspace** — shares this workspace's filesystem and branch,
  so it picks up exactly where this one left off (including uncommitted changes).
- **New workspace** — a fresh, isolated workspace on its own auto-named branch
  cut from the current branch, for continuing the work in a clean checkout.

If `$ARGUMENTS` already makes the choice clear (e.g. it mentions "new agent",
"same workspace", or "new workspace"), use that. **Otherwise, ask the user which
one with your question tool** — `mcp__sculptor__ask_user_question` if it's available, otherwise the built-in `AskUserQuestion` — and
wait for the answer before creating anything. (The tool call raises the
"waiting for input" status that alerts the user; don't ask in plain text.)

## Step 2 — Pre-flight: commit uncommitted work (new-workspace mode only)

**Skip this step if the user chose "new agent, same workspace"** — the same
workspace shares this working tree, so uncommitted work is visible to the new
agent as-is.

For **new workspace** mode, the new workspace is cut from a commit on the
current branch in the project's on-disk git repo. Anything not committed won't
be visible to the new agent. Walk the user through it so they're never
surprised.

Run from the repo root:

```bash
git status --porcelain
```

If the output is non-empty, list the affected paths (just the paths, not the
diffs) and ask with your question tool:

> You have N uncommitted change(s). Commit them before handing off?
> - **Yes, commit them** — write a concise commit message inferred from the
>   conversation and run `git add -A && git commit -m "<msg>"`. Surface the
>   message you used in your next text reply so the user can amend if they want.
> - **No, leave them** — the new agent won't see those changes.

Do **not** prompt for a push. The new workspace shares the user repo's `.git`,
so committed work is visible to it locally — pushing to origin is a separate
concern (CI, MR linkage, code-review visibility) the user can do when they're
ready, and Sculptor's in-workspace diff display doesn't require origin
connectivity either (`origin/*` refs are seeded locally at workspace creation).

## Step 3 — Compose the handoff prompt

Write a self-contained prompt so the new agent can continue without access to
this conversation:

- A summary of the task and the current state of the work.
- Key context: relevant files, decisions made, what is done, what remains, and
  anything already tried that didn't work.
- The next concrete step(s) to take (from `$ARGUMENTS` or the inference above).

Summarize; don't dump the whole transcript.

## Step 4 — Create the handoff

**New agent, same workspace** — the new agent shares this workspace's branch and
files, so reference them by path; it can see them (and uncommitted changes)
directly:

```bash
sculpt agent create \
  -w "$SCULPT_WORKSPACE_ID" \
  --name "<short task name>" \
  --json \
  --prompt "<the prompt from step 3>"
```

**New workspace** — first get the current branch, then create a workspace + agent
based off it:

```bash
git rev-parse --abbrev-ref HEAD   # the source branch

sculpt run \
  --strategy worktree \
  --branch "<current-branch>" \
  --name "<short task name>" \
  --json \
  "<the prompt from step 3>"
```

`--strategy worktree --branch <current>` creates the new workspace on a fresh,
auto-named branch cut from the current branch — the same mechanism the UI uses —
so the committed work carries over.

> **Use `sculpt run`, not `sculpt workspace create`.** `sculpt run` creates the
> workspace, its agent, *and* the git worktree in one step. `sculpt workspace
> create` only registers a workspace record — no agent, no checkout — so it
> renders as a blank screen in the UI. To add an agent to an already-empty
> workspace, run `sculpt agent create -w <ws_id> -p "..."`.

Add `-m sonnet` (or another model) to override the default model.

## Step 5 — Report back

Parse the JSON output and tell the user the new agent ID (and the new workspace
ID, when a new workspace was created) so they can switch to it.

From imbue-ai/sculptor at f847102, shown as plain text under its license: MIT (permissive). The license text is in LICENSE.md.

View SKILL.md on GitHub at f847102

Files

1 file, 5.2 KB in total. Each file has a sha256 in the pin's manifest; the first 12 characters are shown.

Skill folder on GitHub at f847102
All 50 skills from Imbue

More from Imbue

  • architect

    Generates an architecture document for a feature by reading the existing spec, analyzing the codebase deeply, and conducting guided Q&A with the user to…

    Claude Code · about 4K tokens · pinned 6 days ago

  • build

    This skill executes a multi-step implementation plan by working through task files sequentially.

    Claude Code · about 1.1K tokens · pinned 6 days ago

  • build-sculptor-extension

    Build or modify Sculptor extensions, which are runtime ES modules that add panels, overlays, widgets, home views, or settings UI to the Sculptor frontend.

    Contains scripts

    Claude Code, Codex, Cursor +3 · about 3.3K tokens · pinned 6 days ago

  • fix-bug

    Fixes bugs using test-driven development with an interactive or autonomous workflow.

    Claude Code · about 9.8K tokens · pinned 6 days ago

  • help

    This skill fetches the latest Sculptor documentation from GitHub and answers user questions about how to use Sculptor's features, including workspaces…

    Claude Code · about 880 tokens · pinned 6 days ago

  • mock

    Creates interactive HTML mocks for features through guided iteration with users in exploration mode (generating 3+ variants to explore ideas) or confirmation…

    Claude Code · about 3K tokens · pinned 6 days ago

All AI and agents skills
  • agent-builder

    Agent Builder orchestrates the creation, modification, testing, and publishing of n8n Agents.

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +3 · about 2.3K tokens · pinned 6 days ago

  • intent-recognition

    Classifies automation requests by determining which primitive owns the top-level control flow (workflow, agent, needs clarification, or out-of-scope) and…

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +4 · about 6.9K tokens · pinned 6 days ago

  • planned-task-runtime

    This skill manages follow-up interactions with a background task execution system, handling task completion synthesis, failure replanning, workflow building…

    by n8nOfficial publisher

    Claude Code, Codex, Cursor +3 · about 2.2K tokens · pinned 6 days ago

  • skill-creator

    This skill guides agents through creating new skills that extend Gemini CLI with specialized knowledge, workflows, and tool integrations.

    by GoogleOfficial publisher

    Contains scripts

    Gemini CLI · about 5K tokens · pinned 6 days ago

  • -21risk-automation

    This skill automates 21risk operations by connecting to the 21risk toolkit through Composio's Rube MCP server.

    by ComposioOfficial publisher

    Claude Code, Codex, Cursor +3 · about 690 tokens · pinned 6 days ago

  • -2chat-automation

    This skill automates 2chat tasks by integrating with Composio's 2chat toolkit through Rube MCP.

    by ComposioOfficial publisher

    Agents not labelled yet · about 680 tokens · pinned 6 days ago

Open roles at Imbue

No open roles at Imbue on the Agent Native jobs board right now.

Browse the jobs board

The skills listed here are third-party software, collected from public GitHub repositories as a catalog of recent skills and provided as is. Agent Native does not write or control them and is not responsible for any harm, loss or damage caused by installing or using them. Labels and checks describe what we saw at one commit and are not a guarantee. Read a skill's files before you run it. Terms of use