# Handoff

> Write a claimable Vertiso Memory handoff when the user asks to end or transfer a work session, invokes /handoff, or affirmatively accepts an agent's offer before switching tools, ending a session, or approaching compaction. Gather past context through the checkpoint workflow and preserve the handoff's decisions and state without requiring a second capture approval.

- Skill: `vertiso/handoff` (Agent Skill)
- Install (CLI): `npx skillmds@latest add vertiso/handoff`
- Raw SKILL.md: https://api.skillmd.com/api/skills/vertiso/handoff/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: AI & ML
- Author: Vertiso (https://skillmd.com/u/vertiso)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/vertiso/handoff

---


# Handoff

Write a structured session handoff so the work can be picked up cleanly — in this tool or another, now or later.

**Announce:** "Using the handoff skill to save selected session context to Vertiso Memory."

## Consent boundary

Invoking this skill at the user's request, or affirmatively accepting an
agent's offer to run it, authorizes one capture. That capture may inspect the
current conversation and a client-exposed session transcript for exact
quotations, then send selected context and attributed verbatim excerpts to
Vertiso Memory as a claimable handoff. Once invoked, do not ask a second
permission question before performing the transcript-backed handoff write.

One invocation does not grant standing authorization for later handoffs. Omit
and do not persist passwords, API keys, authentication tokens, payment
information, illegal or illicit materials, or sensitive third-party
information the user is not authorized to store.

### Secret stripping

Before composing the memory, inspect every selected source, including the
conversation, transcript, work artifacts, tool output, title, tags, and
metadata, for secrets. Treat passwords, API keys, bearer or refresh tokens,
OAuth codes, session cookies, private keys, recovery codes, and credentialed
connection strings as secrets.

Strip or redact any detected value before composing the memory. Never send the
original value to Vertiso Memory or copy it into a quote, title, tag, metadata,
link, or error detail. Do not echo the secret in the report; state only that
sensitive content was omitted. If sanitization would make the capture
meaningless, stop and ask the user for a sanitized replacement.

## When to use

- The user runs `/handoff`.
- The user asks for a handoff or affirmatively accepts an agent's offer.
- A session ending, a tool or device switch, or approaching compaction is a reason to offer
  a handoff, not authorization to run it. Wait for the user's approval.

## Procedure

1. **Gather the past context via the checkpoint skill.** Follow the `checkpoint` skill's Procedure steps 1-2 —
   gather authoritative state from the workstream's own artifacts (checkpoint's gather step has the domain palette:
   code / writing / product / design / hardware / ops — not git by default) and retrospective-sweep the whole span for
   the attributed verbatim decisions (who said what). This gathered context FILLS handoff's "What happened" and
   "Where it stands" slots **verbatim** — carry checkpoint's attributed quotes and detailed snapshot across intact, do
   not summarize them. It does not replace handoff's body. Do NOT invoke the checkpoint skill itself (that writes a
   checkpoint memory) — reuse its gather + sweep.
2. **Compose handoff's own body** (below), building the forward slots — Next, Open questions, Verify —
   ON TOP of that gathered context. Distinguish what was *done* from what was *told* or *believed*; record an approval
   or decision only if witnessed this session, dated and attributed.
3. **Call the `handoff` verb** — the `handoff` MCP tool, or `vmem handoff` — with the composed body and an
   agent-supplied `scope` (a git branch for a coding session, a project name, or none). This is handoff's terminal act:
   it writes the record AND marks it claimable so `handoff_resume` can pick it up.
4. **Report** the new handoff id to the user.

## The six-slot body

Checkpoint's gather (step 1) fills slots 1-3 with the attributed past context; slots 4-6 are handoff's own, built on it.

1. **Working on** — the subject or goal of the session.
2. **What happened** — what was done and decided, with the attributed verbatim that drove it (from checkpoint's
   sweep).
3. **Where it stands** — the current state in the workstream's terms (from checkpoint's gather).
4. **Next** — the concrete pick-up point.
5. **Open questions / blockers.**
6. **Verify before operating** — the standing advisory: before acting on this handoff, re-check slot 3 against current
   reality and re-confirm any decision in slot 2.

## Authoring honesty

A handoff is trusted only if it is honest. A claim and a verified fact must not look identical. If you did not witness
something this session, say so ("the user said X" / "believed to be Y") rather than asserting it as done.