Machine-readable measurement-and-verification agent skills (supplementary to the CCAI@NeurIPS 2026 paper)

The four machine-readable 'agent skills' that are the central artifact of the paper 'Encoding measurement-and-verification discipline as machine-readable skills so an AI agent enforces it' (Tackling Climate Change with ML @ NeurIPS 2026, Papers track). Each skill is a short structured document an AI analysis agent loads automatically when a task matches its triggering description, encoding the mechanical half of a building-energy research programme's measurement-and-verification (M&V) discipline so the agent enforces it by construction: gold-standard (operational/dev standard), research-standard (epistemic ranking, non-circularity, corrections, honest limits), invariant8-mask (a data loader flagging experiment-contaminated and frozen-logger rows), and dhw-cycle-analysis (domain signal-semantics pitfalls). Methodology documents only; the controller source code and all personal/household data are excluded. CC-BY-4.0.