GPT_Handover_Prompt.md

Christiane Wiemann published a two-step prompt workflow on frauwiemann.de for converting a ChatGPT conversation into a structured Markdown handover document. The first prompt casts the model as a project archivist that extracts 14 sections — including project purpose, key decisions, user preferences, guardrails, open questions and source integrity — while the second prompt runs a quality check on the generated file. The output is intended to be uploaded into a future project or AI environment as compact long-term context.

| | GPT Handover Prompt – Two-Step Version | | | Von Christiane Wiemann \| frauwiemann.de | | | Nutze diese zwei Prompts nacheinander, um aus einem wertvollen ChatGPT-Gespräch | | | eine saubere Markdown-Handover-Datei zu erstellen – bereit für ein neues Projekt | | | oder eine andere KI-Umgebung. | | | --- | | | Schritt 1: Handover erstellen | | | Act as a senior project archivist, knowledge manager, and AI handover editor. | | | Your task is to turn this entire conversation into a high-quality Markdown handover document that can be saved and uploaded into a future project, workspace, or AI system as compact long-term context. | | | The goal is not to create a generic summary. The goal is to preserve the most valuable knowledge from this conversation so that a future AI assistant can continue the work with minimal loss of context, decisions, tone, and strategic intent. | | | Please analyze the full conversation carefully. | | | Create a Markdown document with the following sections: | | | PROJECT HANDOVER SUMMARY | | | 1. Project Purpose | | | Summarize what this project is about, the overall objective, the role the AI has played, and what success looks like. | | | 2. Current Status | | | Capture where the project stands, what has been completed, what is still open, and what is in progress. | | | 3. Important Milestones | | | List the most relevant milestones chronologically. For each: what happened, why it mattered, what changed afterwards. | | | 4. Key Decisions | | | Extract all meaningful decisions. For each: Decision, Rationale, Implication, Status confirmed / provisional / still open . | | | 5. User Preferences and Working Style | | | Capture preferences that materially influence future work: tone, writing style, formatting, level of detail, decision-making style, things the user dislikes. | | | 6. Feedback and Learning | | | Summarize what worked, what did not, corrections made, lessons from previous iterations. | | | 7. Important Context | | | Capture relevant context: stakeholder relationships, organizational context, constraints, dependencies, tools, commercial considerations. | | | 8. Tone, Language and Communication Style | | | Describe the desired communication style: preferred tone, vocabulary, sentence rhythm, formality, use of humour or directness. | | | 9. Explicit Rules and Guardrails | | | List rules that have emerged: what should always be done, what should never be done, what requires verification. | | | 10. Open Questions | | | List unresolved questions, decisions, uncertainties or dependencies. Do not turn them into assumptions. | | | 11. Next Best Actions | | | Suggest the most sensible next steps. Separate agreed next steps from AI-recommended ones. | | | 12. Hidden but Important Patterns | | | Review for patterns not always stated directly but supported by repeated behaviour. Label as "Inferred pattern". | | | 13. What a Future AI Should Remember | | | The 10–20 most important facts, rules, preferences and decisions. Practical and easy to scan. | | | 14. Source Integrity | | | Three lists: Explicitly stated / Strongly inferred / Uncertain – should be verified. | | | OUTPUT REQUIREMENTS: | | | - Clear, structured Markdown | | | - Compact but information-dense | | | - No repeated information across sections | | | - Preserve names, terminology, dates, decisions | | | - Do not invent facts | | | - Flag contradictions rather than silently correcting them | | | - Remove filler and casual conversation | | | - Preserve strategically important nuance | | | - Write so that another AI could continue the work immediately | | | At the end, propose a filename: PROJECTNAME Handover YYYY-MM-DD.md | | | --- | | | Schritt 2: Qualitätscheck direkt danach senden | | | Review the handover summary you just created. | | | Check whether any important decision, correction, stakeholder nuance, recurring user preference, project constraint, or unresolved issue from the original conversation is missing. | | | If something relevant is missing, add it. | | | If anything in the summary is speculative or too confidently phrased, mark it clearly as inferred or uncertain. | | | Then return the final cleaned Markdown version. | | | --- | | | Schritt 2 ist entscheidend. Er holt genau das heraus, was zwischen den Zeilen wichtig geworden ist. |