wtf is open-dossier?

Open Dossier is a Claude skill for building up a persistent, growing body of knowledge from documents, emails, meeting transcripts, notes, and research over time, instead of starting fresh with each new conversation. Most AI conversations lose all their accumulated understanding once the session ends. Open Dossier is built to do the opposite: every document you feed into it adds to a knowledge structure that stays on disk and keeps growing. You start with an empty dossier, a folder on your computer, and add source documents to it as your work progresses. Each source is preserved exactly as it was given, then converted to a readable format, analyzed for what matters, and connected into a set of wiki style pages that combine knowledge from multiple sources. The dossier also keeps a one page running summary of the current state, a log of what was added and when, plans, a list of actions to take, and a record of decisions that were made. A key idea is that every meaningful statement in the dossier is labeled as one of four types: something taken directly from a source, an interpretation or inference made from it, a hypothesis that still needs confirming, or a settled decision. This makes it possible to tell at a glance what is a documented fact versus a guess, and contradictions between sources are flagged rather than silently ignored. You interact with it through a handful of commands: setting up a new dossier, ingesting a document, asking a question that gets answered using everything the dossier knows so far with citations back to sources, checking the dossier's health for stale or contradictory information, or refreshing the summary page. It can also respond to plain language requests like asking it to process a document or answer a question about what was decided. It installs into Claude Code either as a plugin or by copying a skill folder, and can also be uploaded as a skill to Claude.ai or the desktop app. The project is released under the MIT license.