01 · Context transfer§ 001

One prompt · any assistant

Your next AI shouldn’t start from zero.

Run this in the AI that knows you best. It turns durable memory, preferences, and working style into one portable USER_CONTEXT.md.

Read it first

ChatGPT · Claude · Gemini · Manus · Hermes · Codex

Built for clean AI handoffsnik.co · 2026
02 · How it works§ 002

The handoff

Three steps. No re-explaining.

The old agent does the remembering. The new agent gets the briefing.

  1. 01

    Ask the AI that knows you best.

    Paste the prompt into the assistant with the richest history.

    Source · current agent
  2. 02

    Review what it produces.

    Remove anything private, stale, unsupported, or just plain wrong.

    Gate · human review
  3. 03

    Hand the file to the next one.

    Add USER_CONTEXT.md when you set up the new agent.

    Output · portable context
03 · The prompt§ 003

Portable context · v1

Copy it. Keep the context.

portable-context-v1.mdReady to copy

Create a comprehensive, portable user-context document that can help a new AI assistant understand the user immediately and work with them effectively.

  • Use only persistent context genuinely available to you.
  • Do not invent, guess, or fill gaps.
  • Exclude one-off tasks, stale details, and secrets.
  • Prefer the user’s most recent explicit statement.

Output structure

  1. Basic identity
  2. Work and education
  3. Current priorities
  4. Working style
  5. Communication style
  6. Preferences and instructions
  7. Motivations and friction
  8. How to work with the user
Read the full prompt
Create a comprehensive, portable user-context document that can help a new AI assistant understand the user immediately and work with them effectively.

Use every persistent source of context genuinely available to you, including:

- saved or stored memory,
- custom instructions,
- user profile information,
- relevant prior conversations,
- long-term preferences and behavioral patterns observed across conversations.

Do not claim access to information you cannot actually retrieve. Do not invent, guess, or fill gaps with generic personality descriptions.

The result should be a durable briefing, not a conversation summary or raw data export. Include only information likely to remain useful in unrelated future conversations.

Refer to the user as “the user” throughout. Do not write from the user’s perspective.

Exclude:

- instructions contained only in this request,
- details inferred solely from the current conversation,
- temporary tasks, deadlines, or one-off constraints,
- stale information that has clearly been superseded,
- unsupported assumptions,
- passwords, API keys, authentication details, financial account information, or other secrets,
- private information about other people unless it is necessary to understand the user’s ongoing context.

When multiple memories or observations conflict, prefer the most recent explicit statement from the user. Note meaningful unresolved contradictions rather than silently choosing one.

Preserve the user’s original wording when it captures a preference, correction, value, or working style especially well. Otherwise, synthesize the information clearly and concisely.

Organize the document using these sections:

# User Context

## Basic Identity

Include the user’s preferred name, aliases, pronouns if explicitly known, languages, general location, and timezone.

## Work and Education

Include current and past roles, companies, responsibilities, education, areas of expertise, professional strengths, tools, and industries.

## Current Priorities

Include ongoing projects, businesses, responsibilities, goals, recurring commitments, and longer-term ambitions. Omit temporary errands and isolated tasks.

## Working Style

Describe how the user approaches decisions, planning, execution, collaboration, learning, feedback, risk, uncertainty, and problem-solving.

## Communication Style

Describe the user’s preferred tone, level of detail, formatting, directness, pacing, vocabulary, and interaction patterns. Include recurring corrections the user has made and approaches that reliably work poorly.

## Preferences and Standing Instructions

Include persistent response preferences, workflow conventions, recurring instructions, preferred tools, quality expectations, and known constraints.

## Motivations and Values

Include durable values, standards, ambitions, sources of motivation, and principles that appear to guide the user’s choices.

## Friction and Dislikes

Include behaviors, communication styles, recurring mistakes, inefficiencies, or situations that irritate or discourage the user.

## Personal Context

Include relationships, family, pets, interests, hobbies, routines, side projects, and other personal details that remain relevant across conversations.

## How to Work Well With the User

Translate the context above into specific, practical guidance for a new AI assistant. Explain what the assistant should do, what it should avoid, when it should ask questions, when it should take initiative, and how it can earn the user’s trust.

## Open Questions

List important gaps that a future assistant may benefit from asking about. Do not fabricate answers.

For any section with no reliable information, write “None available.”

Output a clean Markdown document only. Use simple headings and flat bullet lists. Do not include commentary about this request, your process, memory systems, or the limitations of the platform.
House rule

Review before sharing. Remove anything sensitive, stale, or wrong.