tttypeone1-ship-it/typeone-ai-skill-extractor

This repository does one narrow thing. It provides a prompt template that distills a long AI chat conversation into a portable Skill you can carry into a brand new conversation window. When a chat with an AI assistant runs long, the assistant gradually builds up a sense of how you think and work, but that understanding usually gets lost the moment you start a fresh window. This template is meant to pull the parts worth keeping out of an old conversation so a new one can pick up faster. The project is intentionally minimal. It is not a chat summarizer and not a collection of many prompts. It holds a single template, kept as clean and useful as possible, so it can be copied and used directly. The template lives at prompts/typeone-skill-extractor.md. The template asks the AI to extract things like the user's stable long term thinking and decision patterns, how the user prefers to collaborate, key moments where the user corrected the AI, facts that are likely to still hold true months later, and any inferences that lack solid evidence and need separate verification. It insists that only patterns get extracted, not one-off events, that every conclusion be backed by the user's own words as evidence, and that the output plainly say when a given window has no relevant information rather than guessing. To use it, you open the template file, copy the whole prompt, and paste it into the old AI conversation window you want to extract from. You then take the output and shape it into a Skill you can bring into a new window, setting aside anything uncertain or unverified in a separate section rather than mixing it into the main result. The README includes a privacy reminder: do not let sensitive material end up in the extracted output, including API keys, tokens, passwords, real names, phone numbers, financial details, internal addresses, or local file paths and usernames. Sensitive evidence should be redacted before being included in the result.