---
title: "Conversation-to-Source Extractor"
type: prompt
description: "Extract reusable public assets from long AI conversations while preserving public/private boundaries."
summary: "Turns a long AI conversation export into separate markdown source files, each classified by sensitivity, so publishable assets are separated from private context."
ai_summary: "Prompt that extracts reusable assets from AI chat exports as classified markdown sources with public/review/private sensitivity labels."
status: public
visibility: public
language: en
topics: [extraction, knowledge-capture]
tags: [workflow, boundary-safe]
audience: [ai-practitioners]
publishedAt: 2026-07-04
updatedAt: 2026-07-04
featured: true
related: [templates/public-private-boundary-checker]
sourceStatus: original
sourceSensitivity: public
useCase: "Mining chat exports for publishable knowledge"
inputRequired: [chat export]
outputFormat: "markdown source files, one per asset"
difficulty: intermediate
copyReady: true
safetyNotes:
  - "Never include client names, internal metrics, or personal data in extracted assets."
  - "Anything ambiguous must be classified as review, not public."
canonicalUrl: "https://wbeen-personal-kb.vercel.app/prompts/conversation-to-source-extractor"
---

## Purpose

Long AI conversations contain publishable assets — insights, workflows,
templates, decisions — buried in private context. This prompt extracts the
assets and leaves the context behind.

## When to use

After a long working session with an AI assistant, before the conversation
scrolls out of memory.

## Inputs

- A chat conversation export (markdown or plain text).

## Prompt

```text
You are a Conversation-to-Source Extractor.

Input: a long AI conversation export.
Task: extract reusable public assets (insights, workflows, templates, decisions)
while preserving public/private boundaries.

Rules:
1. Never include client names, internal metrics, or personal data.
2. Output each asset as a separate markdown block with frontmatter:
   type, title, summary, tags, sensitivity (public | review | private).
3. Flag anything marked "review" for human inspection before publishing.

Output format: markdown source files, one per asset.
```

## Expected output

One markdown block per extracted asset, each with frontmatter declaring its
type and sensitivity.

## Safety notes

- Never include client names, internal metrics, or personal data.
- Anything ambiguous must be classified as `review`, not `public`.

## Related resources

- Template: Public/Private Boundary Checker

---

Canonical page: https://wbeen-personal-kb.vercel.app/prompts/conversation-to-source-extractor
This is the AI-readable Markdown mirror. Public, curated content only.