Wildfire Asset Exposure Risk Dashboard using Claude Agents SDK and Lovable UI
Supercharged weather extremes driven by El Niño increase wildfire ignition probabilities across electrical utility corridors and residential real estate zones in Western North America and Australia. Insurance adjusters and grid operators require real-time risk dashboards that analyze vegetation density, wind speed vectors, and building asset density to estimate fire spread exposure. In this challenge, you will construct a spatial analysis agent using Anthropic's Claude Agents SDK paired with a rapid web interface built using Lovable AI tooling. The Claude agent will utilize extended thinking and tool calling to process GeoJSON property maps, run cellular automata wildfire spread models, and assess asset vulnerability. Your solution will generate interactive threat maps, quantify potential monetary loss per power line segment, and produce risk mitigation strategies directly through a modern web interface.
What you are building
The core problem, expected build, and operating context for this challenge.
Build an interactive wildfire threat assessment agent using Claude Agents SDK and Lovable UI to evaluate spatial power corridor exposure.
How work is evaluated
Evaluates the precision of spatial exposure calculation and agent risk analysis output.
Challenge datasets
Public datasets, plus the private datasets you have access to.
How submissions are scored
These dimensions define what the evaluator checks and which criteria separate a passable run from a strong one.
gis_geometry_validity
Ensure spatial buffer computations are mathematically valid
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
asset_value_accuracy
Percentage accuracy of calculated exposure value • target: 0.98 • range: 0-1
This dimension contributes its full weight only when the submission satisfies the requirement. Partial credit is not awarded.
What you should walk away with
Implement specialized tool use with the Claude Agents SDK.
Build a modern spatial risk visualization prototype with Lovable.
Integrate vector GIS operations (Shapely, GeoPandas) with LLM reasoning.
Simulate physical wildfire spread under varying El Niño wind velocity inputs.
Reference links and supporting material
Synthetic GeoJSON layers representing electrical power grid transmission lines, building footprints, and surface wind vector grids.
Challenge setup
Evaluates the precision of spatial exposure calculation and agent risk analysis output.
Challenge input
GeoJSON containing powerline geometry and regional wind vectors
Claude Agents SDK
Mandatory SDK for Claude reasoning and structured agent tools.
Lovable
Rapid UI dashboard rendering and frontend scaffolding tool.
Evaluated output
JSON with total_asset_value_usd, high_risk_structures_count, risk_level
- Ensure spatial buffer computations are mathematically valid
- Percentage accuracy of calculated exposure value • target: 0.98 • range: 0-1
- Benchmark: GIS-ExposureBench
- Asset Value Accuracy target: 0.98
- 1 public reference case
- Python execution harness
- Python sandbox (unavailable on Versalist)
Summary shows the first task, up to three tools, and up to three evidence items. Open the technical recipe to view all configured items.
View technical recipe
Configured tools
- Lovable · Required
- Anthropic · Optional
- RunPod · Optional
Evaluation contract
- gis_geometry_validity · Weight 1
- asset_value_accuracy · Weight 1
Metrics
- asset_value_accuracy · Target 0.98
Benchmarks
- GIS-ExposureBench
- security
Reference cases
- 1 public reference case
Workflow tasks
- Task 1
Input: GeoJSON containing powerline geometry and regional wind vectors
Output: JSON with total_asset_value_usd, high_risk_structures_count, risk_level
Execution harness: Python
Evaluation tools
- Claude Agents SDK · Agent System
- Lovable · AI Engineering Tool
Recipe freeze marker
No recipe freeze time is recorded.