fsilavong

Evaluate agentic systems at component and end-to-end level, producing a structured report with metrics, datasets, and regression strategy. Use this skill whenever the user wants to evaluate, benchmark, or test any agentic or AI pipeline system, including RAG pipelines, multi-agent systems, LLM chains, retrieval systems, or any system with modelling components. Triggers include: "evaluate my agent", "benchmark my pipeline", "write evals for", "test my RAG", "how is my system performing", and "set up evals". Always use this skill when the user shares a codebase and asks how well it works.

Use this skill when an agent needs to write documents in the Unfoldoc format for Orbit or other systems. It teaches the required folder shape, how to write `content.md`, how to link citations and local source files, and when to avoid execution-oriented metadata. Prefer this skill whenever the goal is to produce an Unfoldoc-compatible markdown bundle that the Unfoldoc UI can render.