Yogendra Miraje describes FactSet's move from proprietary blueprints to an open skillsAn AI agent skill is a reusable package of instructions, resources, and tool guidance for performing a bounded kind of work. format. He separates prompts, which define an agent's role, from toolsAI tool calling lets a model request a structured operation from software instead of only returning natural-language text., which determine its connections, and skills, which describe how tasks should be completed.
His minimal harnessAn AI agent harness is the software framework that packages a model with tools, instructions, context management, execution controls, and user interaction. combines a registry, system prompt, file-reading tool and agent loop. A financial-research example demonstrates on-demand instructions and distinct HTML or PDF routing. Miraje recommends organizing skills around user intent rather than underlying datasets.
A model upgrade broke unchanged skills because important instructions appeared late in the file. That example motivates model-specific evaluationEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria., named maintainers, tool access boundaries and versioning. For larger libraries, he discusses retrieval and hierarchy while framing governanceGovernance defines decision-making authority, rules and accountability within a system or organization. as practical maintenance rather than unnecessary bureaucracy.
Watch on YouTube




