---
name: handover
description: "Generates a structured handover document at the end of a chat, or when context is filling up, so the next chat, project, or AI platform can pick up cleanly without re-briefing. Use this skill whenever the user says \"handover\", \"handoff\", \"wrap this up\", \"summarise for next time\", \"context is getting full\", \"I want to continue this in a new chat\", \"give me something I can paste into the next session\", or whenever a long, dense chat contains decisions, reasoning, or open threads worth preserving. Apply proactively when a session has produced material that would be lost if not captured. Output is a single, paste-ready handover block, structured for fast scanning. This is a working brief for the next session, not a chat summary."
---

# Context Handover

A Bolt Method Skill for capturing the working state of a chat so the next session, project, or AI platform can continue the work immediately, with the relevant context and any missing sources clearly identified.

## 1. Purpose

Long chats produce real intellectual capital; decisions taken, the reasoning behind them, options weighed and rejected, threads still open. A later conversation may not have access to those decisions or their context. Capture what matters deliberately.

This skill produces a single, structured handover document the user can:

1. Paste into a new chat in the same project or a different one.
2. Use in Claude or ChatGPT to continue the same line of work.
3. File as a session record.
4. Use as a memory anchor when returning to a workstream after time away.

The handover is not a transcript or a chat summary. It is a working brief the next instance of the assistant can operate from immediately.

This skill captures the working state of a single session. Keep standing preferences separate from this session record, and check what the next conversation can access.

## 2. When to trigger

Generate a handover when any of the following are true:

1. The user explicitly asks ("handover please", "wrap this", "give me a handover block").
2. The chat is visibly long and dense, and the user signals they are stepping away or switching context.
3. A meaningful decision has been reached and the user mentions continuing the work elsewhere.
4. The user mentions context filling up, the chat slowing, or wanting to start fresh.
5. The work has crossed more than one area and needs to land in more than one place.

Do not generate a handover for short, transactional chats. The threshold is roughly: would the next session need more than two minutes of re-briefing to be useful? If yes, a handover is warranted. If the user asks for one but the chat is genuinely thin, say so directly and offer a three-line note instead.

## 3. The handover structure

Every handover follows this eight-section structure. Keep the shape consistent; the value is partly in the predictability, so the reader knows where to look.

Wrap the entire handover in a single fenced code block so it copies cleanly in one action.

```
HANDOVER: [One-line title for this handover]
Date: [DD Month YYYY]
Source chat: [One-line description of what this chat was about]
Target: [Where this is going next; e.g. "New chat, same project", "ChatGPT for a second opinion", "Ops chat"]

1. ONE-LINE STATE
[Single sentence. Where things stand right now. Nothing else.]

2. DECISIONS TAKEN
[Numbered list. What was decided in this chat, with the reasoning compressed to one line each. Include only settled decisions, not options still under consideration.]

3. KEY REASONING
[The "why" behind the decisions above. This is the load-bearing section; it stops the next session re-litigating settled ground. Two to five short paragraphs. Include trade-offs considered and rejected.]

4. CURRENT STATE OF PLAY
[Numbered list. What exists now that did not exist at the start of the chat. Documents drafted, frameworks built, numbers agreed, wording locked. Include file names or references where relevant.]

5. OPEN THREADS
[Numbered list. What is still unresolved. For each: the question, what is needed to close it, and what or who is the blocker. Be honest about threads opened and parked rather than closed.]

6. NEXT ACTIONS
[Numbered list, in priority order. Concrete, specific, time-bound where possible. Distinguish the user's own actions from actions for other people, and from actions the next session should take on first contact.]

7. CONTEXT THE NEXT SESSION NEEDS
[Anything not already in the project instructions or persistent memory that the next session must know to be useful when the next conversation starts. Names with roles, key numbers, wording that has been locked, sensitivities, recent shifts in position. Keep it tight; do not repeat what is already stored elsewhere.]

8. WHAT NOT TO REOPEN
[Numbered list. Decisions, options, or framings considered and rejected in this chat. This protects the next session from helpfully resuggesting things already worked through and dismissed. If nothing was rejected, say so explicitly.]
```

## 4. Writing rules

1. Be brief and direct. No filler, no preamble, no closing pleasantries.
2. Each list item is one to two lines maximum. The Key Reasoning section is the only place for paragraphs.
3. Use numbered lists, not bullets, in every list section.
4. Name people with their role on first mention inside the handover (e.g. "Andy, operations lead") so a fresh reader understands the cast.
5. Use figures for money, percentages, dates, and quantities above ten.
6. Never include sensitive items that should not be re-pasted (full bank details, passwords, login credentials, detailed personal or medical information). Reference them obliquely if relevant to the work.
7. Match the user's own language conventions if they have stated any (spelling variant, punctuation preferences, capitalised terms). If the user has a framework or house terminology, use it; otherwise stay neutral.

## 5. Variants

The user may ask for a variant of the standard handover. Three are supported:

5.1 Lean handover ("quick handover", "short version")
Sections 1, 2, 5, and 6 only. Used for shorter chats, or when the work is mostly settled and only a few threads remain.

5.2 Cross-platform handover ("for ChatGPT", "for Claude")
Standard structure, plus a final section, 9. PLATFORM NOTE, with one line on what that specific platform is being asked to do (e.g. "Stress-test the financial assumptions in section 4"). Strip any references specific to the current assistant so the block reads cleanly elsewhere.

5.3 Filed handover ("file this", "save it", "for the project folder")
Standard structure, saved as a downloadable .md file. Suggested naming convention:
HANDOVER_[YYYY-MM-DD]_[three-word-topic-slug].md
For example: HANDOVER_2026-06-07_pricing-model-review.md
If the user asks to file or save the handover, write it to the outputs directory and present it so they can drop it straight into the relevant folder.

## 6. Quality check before delivery

Before producing the handover, run this internal check:

1. Could a fresh session, reading only this handover plus any project instructions, pick up the work with the essential context and any missing sources clearly identified? If not, what is missing?
2. Is the Key Reasoning section actually load-bearing, or just restating the decisions? It should explain why, not what.
3. Are open threads honestly labelled as open, or have parked items been quietly promoted to "next actions" to make the handover feel tidier? Do not do this.
4. Does the What Not To Reopen section contain real items, or an explicit note that nothing was rejected?
5. Is the language clean and consistent with any conventions the user has stated?

## 7. Edge cases

The chat is mostly emotional or wellbeing-related, not work. Do not produce a standard handover. Offer a short, private note instead, and ask the user where, if anywhere, they want it stored. Default to not storing.

The chat involves sensitive commercial, legal, or personal matters. Flag in section 7 if anything is sensitive enough to warrant not pasting into a general working chat or another platform.

The chat is genuinely thin. Say so directly: "There is not enough in this chat to warrant a full handover. Do you want a three-line note instead?"

## 8. Delivery

When triggered, lead with one short sentence naming what the handover is for, then the fenced block. No further commentary unless the user asks.

Example opener:
"Handover for the pricing model session, ready to paste into a new chat:"

Then the code block. That is it.
