What problem does it solve? Running geophysical simulations and inversions with SimPEG requires operational expertise scattered across source code, docs, and examples: correct SI unit conventions, mesh design, map selection, regularization tuning, and recovery from silent failures. This Skill packages that expertise so an AI agent can reliably build meshes, configure models, execute forward simulations and inversions, parse results, and diagnose failures without substituting approximations for the real model. ## Core Features & Use Cases - Staged pipeline tools: Four executable scripts cover mesh and survey setup (build_mesh.py), model initialization with unit-aware maps (initialize_model.py), forward/inversion execution (run_simpeg.py), and output parsing with convergence metrics and plots (parse_results.py). - Diagnostic recovery: 18 symptom-to-diagnosis-to-remedy triplets in diagnostics/triplets.yaml cover known failure modes such as CGS-vs-SI density errors, resistivity/conductivity confusion, singular air-cell conductivity, and beta mis-scheduling. - Machine-readable contracts: dag.yaml, format_spec.yaml, and validation_convention.yaml define inputs, outputs, units, observability, and validation metrics so runs are graded against cited conventions rather than intuition. - Use Case: A user provides gravity survey locations and observed data in CSV; the agent builds a padded TensorMesh, initializes a density model in kg/m^3, runs a beta-cooled Gauss-Newton inversion, and exports the recovered model CSV with convergence diagnostics. ## Quick Start Ask the agent to run preflight_check.py first, then build a mesh and run a gravity inversion from your survey CSV using the SimPEG pipeline tools.
