The Agent Skills Directory

[DYNAMIC_EXECUTION]: The skill provides implementation examples using pickle.loads() to retrieve data from context.job_cache. As pickle is inherently unsafe for deserialization, this pattern introduces a risk of arbitrary code execution if the cache content is manipulated or contains untrusted data.
[EXTERNAL_DOWNLOADS]: The documentation guides users to manage dependencies using uv sync and pyproject.toml for various well-known geospatial and machine learning libraries, including rasterio, zarr, xarray, odc-geo, and torch.
[DATA_EXFILTRATION]: The skill describes procedures for interfacing with established cloud storage services (Amazon S3, Google Cloud Storage, and Azure Blob Storage) using the obstore library for high-throughput geospatial data operations.
[PROMPT_INJECTION]: The skill provides patterns for processing external geospatial datasets which introduces a surface for indirect prompt injection. 1. Ingestion points: dataset.query and GeoTIFF.open operations. 2. Boundary markers: Not present in the instructional patterns. 3. Capability inventory: Shell command execution via uv and network operations via niquests and obstore. 4. Sanitization: No explicit data validation or sanitization steps are included in the provided workflow patterns.

writing-tilebox-workflows