Skilldocs might finally kill the Notion bloat for Markdown lovers

Notion has become an absolute labyrinth of nested databases and proprietary blocks that make it feel more like a bloated CRM than a lightweight writing tool. I've been looking for a way to treat documentation more like a high-fidelity design file, and this new project Skilldocs seems to be hitting that specific niche by treating Markdown with the same collaborative intensity that Figma applies to vector graphics.

The core premise is simple but effective: it's a collaborative Markdown editor that brings real-time multiplayer features—the kind you expect from Google Docs or Figma—into a pure Markdown environment. If you are building complex AI workflows or managing detailed LLM agent documentation, having a structured, plain-text-first environment is non-negotiable.

Why a Markdown-first approach matters for AI workflows

When we are designing agentic workflows or defining "skills" for an LLM, we don't need heavy UI widgets. We need structured, readable, and portable text. The biggest friction point with tools like Notion is the "export struggle." You spend hours perfecting a page, only to have the formatting break when you try to feed that content into a RAG (Retrieval-Augmented Generation) pipeline or a prompt engineering template.

Skilldocs solves this by making Markdown the source of truth. Because it's native, you can actually leverage the collaborative features without the proprietary metadata baggage.

The "Figma for Docs" feature set

What actually caught my eye isn't just the syntax highlighting, but the specific way it handles real-time collaboration:

• Multiplayer Cursor Tracking: You can see exactly where your teammate is looking or typing, which is indispensable during live technical syncs or pair-programming sessions.
• Live Following: Much like Figma, you can "follow" a collaborator. If you are on a call explaining a complex technical spec, you can jump to their view instantly.
• Comment & Diff Export: This is the killer feature for anyone working with AI agents. You can export the discussion threads and the diffs directly to an agent. Instead of just getting the final text, you can feed the context of why changes were made into your LLM.

A practical way to integrate this into an AI workflow

If you are building a documentation-heavy project, I’d suggest using this as a staging area for your agent instructions. Here is a conceptual way to think about the deployment of your documentation:

1. Drafting: Use the real-time editor to live-edit skill definitions or system prompts with your team.
2. Reviewing: Use the comment system to debate specific constraints or logic gates within the Markdown.
3. Agent Integration: Instead of a manual copy-paste, export the finalized Markdown along with the "reasoning" (the comments) to your LLM agent to provide high-context fine-tuning or prompt updates.

It feels like we are moving away from "all-in-one" productivity suites and toward specialized, interoperable tools. If you spend your day in a terminal or a code editor, a tool that respects Markdown and actually plays nice with other developer tools is a massive upgrade over the "walled garden" approach of modern workspace apps.