quantiota/ska-protocol — explained in plain English
Analysis updated 2026-05-18
Coordinate multiple autonomous AI agents so they build shared knowledge without a central controller.
Log and persist human-agent conversations into a database for later context retrieval.
Study a proposed protocol for structured knowledge accumulation before implementing it.
| quantiota/ska-protocol | andybowu/hitchhikers-guide-agentic-ai-zh-cn | harshkaso/resumetex | |
|---|---|---|---|
| Stars | 3 | 2 | 4 |
| Language | TeX | TeX | TeX |
| Setup difficulty | — | easy | easy |
| Complexity | — | 1/5 | 2/5 |
| Audience | researcher | researcher | general |
Figures from each repo's GitHub metadata at analysis time.
SKA Protocol is a project describing how a group of AI agents can work together to build and accumulate shared knowledge, both among themselves and from their conversations with humans. According to its README, the project has grown too large and too broad in scope for one person, or a single automated agent, to build alone, so it is being developed by a coordinated group of eight separate agents instead of one. The framework the project is built around covers three main areas. The first is agent to agent knowledge sharing: agents send messages to each other, pull out useful pieces of knowledge from what they learn, track how confident they are in that knowledge, notify each other of new findings, and build up a shared body of knowledge without any single agent directing the others. The second area is human to agent knowledge, which covers logging conversations from a terminal, storing that information in a database called QuestDB, and letting an agent retrieve relevant past context and pass memory along even if a different agent picks up the conversation later. The third area is the production infrastructure needed to support all of this: streaming data in real time, ingesting it in batches, handling failures gracefully, and managing timestamps correctly. To manage this scope, the project describes an eight node setup, where each node is a separate agent responsible for one part of the system, such as the core protocol rules, the messaging layer, the database schema, or security and documentation review. These eight agents are described as independent peers rather than workers taking orders from a central controller: each one can focus on its own area, share findings, work on its own code branch, and check another agent's work, though any of them can flag a problem to the human who created the project. That human author's role shifts from writing every piece of code personally to setting the overall principles, approving major decisions, and acting as the final authority on what the underlying idea, called Structured Knowledge Accumulation, actually means. Before writing any new code, the stated first task for the eight agents is simply to study the existing project together, map out what already exists, and agree on what to build next, so that the group does not end up building eight different, incompatible versions of the same system.
SKA Protocol coordinates eight independent AI agents to build a shared system for accumulating agent-to-agent and human-to-agent knowledge over time.
Mainly TeX. The stack also includes QuestDB.
Mainly researcher.
This repo across BitVibe Labs
double-check against the repo, no cap.