Using Geoscience Skills

Using Geoscience Skills

Meta-skill for discovering, routing, and composing the geoscience skills library. This skill maps user intent to domain skills, workflow skills, slash commands, and agents.

Domain Routing Table

Match user intent keywords to the appropriate domain skill.

Workflow Skills

Workflow skills chain multiple domain skills into end-to-end pipelines.

Available Agents

All 29 Domain Skills by Category

Seismic and Seismology

• segyio -- SEG-Y file I/O, trace and header access
• obspy -- seismological waveform processing, FDSN services
• disba -- surface wave dispersion (Rayleigh, Love)

Well Log Analysis

• lasio -- LAS file reading and writing
• dlisio -- DLIS/RP66 binary well log parsing
• welly -- well data analysis, curve QC, multi-well projects
• petropy -- petrophysical analysis, formation evaluation
• striplog -- lithological and stratigraphic log display

3D Geological Modelling

• gempy -- implicit 3D geological modelling
• loopstructural -- 3D modelling with fold and fault support
• gemgis -- spatial data preprocessing for GemPy

Geophysical Inversion

• simpeg -- multi-method geophysical inversion framework
• pygimli -- ERT, SRT, IP inversion with simple API
• devito -- symbolic PDE solver for wave propagation
• pylops -- linear operators for inverse problems

Potential Fields and Rock Physics

• harmonica -- gravity and magnetic data processing
• bruges -- AVO, Gassmann, wavelets, elastic moduli

Spatial Analysis and Geostatistics

• verde -- spatial gridding and interpolation
• geostatspy -- variograms, kriging (GSLIB-style)
• scikit-gstat -- geostatistics with scikit-learn API

Hydrology and Surface Processes

• pastas -- groundwater time series modelling
• landlab -- landscape evolution modelling

Structural Geology and Geochemistry

• mplstereonet -- stereonet plots for orientation data
• pyrolite -- geochemical analysis and diagrams

Near-Surface Geophysics

• gprpy -- GPR data processing
• mtpy -- magnetotelluric data analysis

Data Formats and Visualization

• xarray -- NetCDF, multi-dimensional labeled arrays
• pyvista -- 3D mesh visualization and analysis
• pooch -- data file fetching and caching

Skill Composition Rules

Chain skills when a task spans multiple stages of a geoscience workflow.

Composition Patterns

Data Loading -> Processing -> Modelling -> Visualization

1. Always start with a data I/O skill (segyio, lasio, dlisio, xarray)
2. Use processing skills for QC and transformation (welly, obspy, verde)
3. Apply domain modelling (bruges, gempy, simpeg, pygimli)
4. Finish with visualization (pyvista, matplotlib via domain skill)

When to Chain vs Use Standalone

Dependency Awareness

When composing skills, respect data flow:

# Correct: segyio loads, obspy processes, bruges models
import segyio
import obspy
from bruges.reflection import zoeppritz

# Load with segyio
with segyio.open('seismic.sgy') as f:
    data = f.trace[:]

# Process with obspy (convert to Stream if needed)
# Model with bruges
Rpp = zoeppritz(vp1, vs1, rho1, vp2, vs2, rho2, theta)

When to Use This Skill

This is the discovery and routing skill. Use it when:

• Starting a new geoscience coding session and unsure which library to use
• A user request spans multiple geoscience domains
• You need to find the right slash command or workflow for a task
• Composing multiple domain skills into a pipeline
• Looking up which skill handles a specific file format or analysis type

This skill does not perform any analysis itself. It directs to the appropriate domain skill, workflow skill, or agent for execution.

Quick Decision Tree

User wants to...
  |
  +-- Load/write a file? --> Check format:
  |     SEG-Y -> segyio    LAS -> lasio    DLIS -> dlisio
  |     NetCDF -> xarray    VTK -> pyvista
  |
  +-- Process signals? --> obspy (seismology), welly (well logs)
  |
  +-- Build a model?
  |     3D geology -> gempy or loopstructural
  |     Rock physics -> bruges
  |     Inversion -> simpeg or pygimli
  |
  +-- Visualize results? --> pyvista (3D), matplotlib (2D via domain skill)
  |
  +-- Run a full pipeline? --> Use workflow skills above