How do I install the Raster At Scale skill?

Run npx skillmds@latest add buildmoonshot/raster-at-scale in your terminal (requires Node.js), paste this page's agent-chat prompt into Claude, Cursor, or any MCP-connected agent, or download the SKILL.md file and copy it into your agent's skills directory.

Is Raster At Scale safe to use?

This skill has not completed SkillMD's automated safety review yet. SkillMD never runs a skill's scripts for you; review the SKILL.md before installing.

Which AI agents work with Raster At Scale?

This skill is tagged as working with Claude Code, Claude.ai, OpenAI Codex. SKILL.md is an open format, so most agents that read a skills directory can load it too.

Is Raster At Scale free to use?

Yes. Installing skills from SkillMD is free, and the skill stays under its author's original license.

Who published Raster At Scale?

buildmoonshot (@buildmoonshot) published this skill. Their other Agent Skills are listed on their SkillMD profile.

Raster At Scale by buildmoonshot — Agent Skill

Name: Raster At Scale
Author: buildmoonshot

Raster At Scale

Use when processing raster or imagery data that is large — multi-gigabyte GeoTIFFs, mosaics, satellite scenes, DEMs — where loading the whole array into memory would fail or thrash. Makes the agent work in windows/tiles, use overviews and cloud-optimized formats, and avoid out-of-memory full-array reads.

buildmoonshot 4f89e0c 2 files · 3.5 KB Updated Sep 17, 2026

File contents

Raster at Scale

A large raster will not fit in memory, and the naive read everything into a NumPy array is how raster jobs crash. Process big rasters the way they're built to be processed.

Don't load the whole thing

Read in windows/blocks, not all at once (rasterio windowed reads, GDAL block reads). Process tile by tile and write tile by tile.
Read in the raster's native block size where you can — aligned reads are far faster than arbitrary windows.

Use the right format and pyramids

Prefer Cloud-Optimized GeoTIFF (COG) for efficient partial/range reads.
Build overviews (gdaladdo) so display and downsampled analysis don't touch full resolution.
For chunked array work, reach for xarray/dask or rioxarray rather than a monolithic array.

Reproject and warp without loading

Resample/reproject with gdalwarp (use -multi -wo NUM_THREADS=ALL_CPUS), not by reading arrays into Python.
Pick the resampling method by data type: nearest for categorical/classified rasters (preserves class values), bilinear/cubic for continuous (elevation, imagery). Using bilinear on a land-cover raster invents classes that don't exist.

Get the details right

Honor the NoData value so it doesn't poison statistics or show up as real zeros.
Confirm the CRS before warping or aligning to other layers (see crs-discipline).

Why this matters

The failure modes at scale are distinct: out-of-memory crashes, hour-long full-resolution passes that should have used overviews, and silently wrong resampling that corrupts categorical data. Windowed reads, COGs/overviews, and method-appropriate resampling turn an unworkable raster job into a fast, correct one.

---name: raster-at-scaledescription: Use when processing raster or imagery data that is large — multi-gigabyte GeoTIFFs, mosaics, satellite scenes, DEMs — where loading the whole array into memory would fail or thrash. Makes the agent work in windows/tiles, use overviews and cloud-optimized formats, and avoid out-of-memory full-array reads.---# Raster at ScaleA large raster will not fit in memory, and the naive `read everything into a NumPy array` is how raster jobs crash. Process big rasters the way they're built to be processed.## Don't load the whole thing- Read in **windows/blocks**, not all at once (`rasterio` windowed reads, GDAL block reads). Process tile by tile and write tile by tile.- Read in the raster's **native block size** where you can — aligned reads are far faster than arbitrary windows.## Use the right format and pyramids- Prefer **Cloud-Optimized GeoTIFF (COG)** for efficient partial/range reads.- Build **overviews** (`gdaladdo`) so display and downsampled analysis don't touch full resolution.- For chunked array work, reach for `xarray`/`dask` or `rioxarray` rather than a monolithic array.## Reproject and warp without loading- Resample/reproject with **`gdalwarp`** (use `-multi -wo NUM_THREADS=ALL_CPUS`), not by reading arrays into Python.- Pick the resampling method by data type: **nearest** for categorical/classified rasters (preserves class values), **bilinear/cubic** for continuous (elevation, imagery). Using bilinear on a land-cover raster invents classes that don't exist.## Get the details right- Honor the **NoData** value so it doesn't poison statistics or show up as real zeros.- Confirm the CRS before warping or aligning to other layers (see `crs-discipline`).## Why this mattersThe failure modes at scale are distinct: out-of-memory crashes, hour-long full-resolution passes that should have used overviews, and silently wrong resampling that corrupts categorical data. Windowed reads, COGs/overviews, and method-appropriate resampling turn an unworkable raster job into a fast, correct one.

Frequently asked questions