Build a STAC Wildfire Defensible Space Compliance Auditor — implementation

## Build
Discover SpatioTemporal Asset Catalog imagery items over commercial property boundaries to evaluate wildfire fuel buffers. Fetch imagery assets to identify vegetation encroachment zones and construct actionable wildland-urban interface mitigation plans.

## User
Wildfire Mitigation Planner

## Workflow
1. Parse property footprint GeoJSON features and perimeter buffer zones.
2. Retrieve multispectral STAC Items intersecting the spatial boundary and fetch the linked imagery assets.
3. Calculate vegetative fuel encroachment inside the 30-meter defensible perimeter from the downloaded imagery.
4. Produce an adaptation action plan with vegetative thinning priorities.

## Interface
Interactive map viewer highlighting buffer zones with side-by-side satellite asset comparisons and adaptation action tickets.

## Data
Query public STAC API collections using sample bounding boxes. Use mock GeoJSON property boundaries in test directory.
Starter examples below are synthetic. They are not downloaded records.

## Source
[STAC specification](https://raw.githubusercontent.com/radiantearth/stac-spec/master/README.md)
Consumes STAC Item specification GeoJSON Feature attributes, foreign members, and linked asset objects to access and download raster imagery scenes.

## Tools
dspy, pystac, rasterio.
Optimizes structured extraction prompts that evaluate satellite metadata against defensible fuel criteria.
Requires an LLM supporting structured generation pipelines with prompt optimization modules.

## Deliverables
Submit runnable source, pinned dependencies, setup commands, a working interface, example outputs, and a recorded end-to-end demo. Include tests and an action log.

## Acceptance
- Extracts bounding box coordinates conforming to STAC Item core specification.
- Downloads linked multispectral imagery assets to evaluate vegetation cover.
- Generates vegetation thinning priority rankings with quantitative buffer distances.
- Produces valid GeoJSON adaptation polygons for property owner review.

## Starter Examples
```json
[
  {
    "input": {
      "property_id": "PARCEL-WUI-404",
      "stac_item_id": "S2B_MSIL2A_20230715",
      "buffer_radius_m": 30,
      "vegetation_cover_pct": 58.4,
      "compliance_threshold_pct": 30,
      "record_type": "simplified synthetic app record"
    },
    "expected_output": {
      "property_id": "PARCEL-WUI-404",
      "defensible_space_compliance": "NON_COMPLIANT",
      "recommended_reduction_pct": 28.4,
      "adaptation_urgency": "HIGH",
      "action_code": "THIN_CANOPY_30M"
    }
  },
  {
    "input": {
      "property_id": "PARCEL-WUI-405",
      "stac_item_id": "S2B_MSIL2A_20230718",
      "buffer_radius_m": 30,
      "vegetation_cover_pct": 14.1,
      "compliance_threshold_pct": 30,
      "record_type": "simplified synthetic app record"
    },
    "expected_output": {
      "property_id": "PARCEL-WUI-405",
      "defensible_space_compliance": "COMPLIANT",
      "recommended_reduction_pct": 0,
      "adaptation_urgency": "NONE",
      "action_code": "MAINTAIN_CLEARANCE"
    }
  }
]
```

## Safety
Generates simulated vegetative thinning recommendations; does not guarantee physical wildfire defensibility. All external actions remain mock or human-reviewed.
Use mock external writes. Require human review before any real deployment.