AI product
docume is a framework for comparing and evaluating agent-related components and context, including skills, AGENTS.md files, and MCP tools. It splits a skill into small, checkable claims, tests a model with and without each claim, and produces an initial vector showing what the model already knew, what the skill teaches, and which details may make performance worse. The framework also covers evaluations, context-impact assessment, routing, and comparison of descriptions and naming.
1 use taken from transcripts — each links to the moment in the video.
A framework for comparing and evaluating everything related to agents and context — skills, AGENTS.md, MCP tools, and more. Its novelty feature splits a skill into checkable items and gives an initial vector for skill optimization, alongside evals, context impact assessment, routing, and comparing descriptions and naming.
1 in the library.