---

name: features-dataframes-analysis description: "Work with feature layers, FeatureSet/Feature, spatially enabled dataframes, geometry operations, replicas, branch versioning, and hosted feature analysis." disable-model-invocation: true metadata: disco-role: operating license: Apache 2.0

Features, DataFrames, and Analysis

Use this sub-skill when the task is about:

• feature layers, tables, Feature, FeatureSet, and FeatureLayerCollection
• spatially enabled dataframes (SEDF), spatial joins/overlays, plotting, and local vector IO
• Geometry creation, validation, projection, and spatial predicates
• edits, appends, attachments, replicas, and branch versioning
• hosted feature and spatial analysis functions in analysis, manage_data, use_proximity, find_locations, and summarize_data

Start here:

• features and SEDF workflows
• geometry and analysis API
• troubleshooting
• local geometry smoke

Operating rules:

1. Separate local geometry/SEDF work from hosted service calls.
2. Before editing, appending, versioning, or analyzing a layer, validate field names and types, object ID/global ID availability, geometry type, spatial reference, and service capabilities.
3. Use query(return_count_only=True) and query(return_ids_only=True) as preflight checks before pulling large result sets.
4. Use FeatureLayer.append for bulk loads; do not assume a public manage_data.append_data helper exists.
5. For local data, prefer GeoAccessor.from_layer, from_featureclass, from_xy, to_featureclass, to_featureset, and to_featurelayer when the task stays on disk or in memory.
6. For hosted analysis functions, expect service-backed outputs, async jobs, and possible credit usage.
7. For sync-enabled or branch-versioned services, confirm syncEnabled or hasVersionedData before using replicas or versions.
8. Use the smoke script before writing a service-dependent workflow.

Route elsewhere when the request is about:

• imagery or raster analytics -> imagery-raster-analysis
• geocoding, routing/network, geoenrichment, or map widgets -> mapping-location-services
• arcgis.learn or model training/inference -> deep-learning