What is fundemetal?

Fundemetal is a small collection of thinking frameworks meant to be used with AI assistants like Claude, TRAE, Cursor, or Windsurf. It is not a piece of software you run, it is a set of plain text markdown files that you paste into an AI agent's system prompt or drop into its skills folder, so the agent follows a specific way of thinking before it acts. The project includes three related protocols. The first, called Decomposer, is meant to help surface the things you do not even know you do not know about a problem, turning vague uncertainty into a concrete list of open questions. The second, called Tension Mining, looks for the underlying conflicting forces at play in a system and tries to find patterns that repeat across different fields. The third, Great Expectations, is aimed at planning in a way that avoids simply copying whatever the obvious or most common answer would be, and includes checks meant to catch weak or incomplete plans. Each of these three files shares the same design rules. They are meant to enforce a genuine change in how the AI reasons rather than just performing a single task, they include built in checks meant to catch shallow or dishonest answers, they are based on established thinking frameworks from fields like psychology and complexity theory, and they work with any AI agent since they are just markdown text with no code or installation required. The suggested way to use them is in sequence: first Decomposer to expose the real unknowns, then Tension Mining to understand the deeper forces at work, and finally Great Expectations to turn that understanding into an actual plan. The project is released under the MIT license, so it can be freely copied, modified, and shared. It would suit people who use AI assistants for planning or problem solving and want more rigorous, less generic output from those tools, though results depend entirely on how well the AI agent follows the instructions given to it.