kimyx0207/kim_decision

This repository has been merged into Kim Service and no longer publishes standalone releases. Please use Kim Service for new versions, installation instructions, and release notes.

Overview

KIM Skill is a decision protocol for turning fuzzy questions into usable next actions.

It is built for the moments where a normal prompt usually produces confident but vague advice: should I launch, sell, price, cut scope, change direction, or run a test first? KIM forces the agent to state the intent, map the path, label the evidence, define the smallest useful test, and end with something you can execute.

Most AI skills tell the model what style to use. KIM asks a different question: does the agent actually have a method for reaching a usable result — or is it just improvising with confidence?

The point is simple: less inspirational advice, more decisions that can be tested.

Project info

This repository is the public source for the KIM decision protocol. The GitHub README is the user-facing entry point; the skill directories are the runtime packages; docs/zh-CN/ contains the Chinese reference materials used by the skill.

Without a method, the model improvises. With KIM, every output follows one spine:

flowchart LR
    subgraph define["Define"]
        I["IntentWhat must change"] --> S["SubjectWho experiences it"]
        S --> P["PathHow they get there"]
    end

    subgraph verify["Verify"]
        C["ConstraintHard limits"]