The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill documentation and example scripts specify the use of standard Python packages such as deweypy, duckdb, and pandas. It also references an R package deweyr hosted on GitHub for R-based workflows.
[COMMAND_EXECUTION]: The skill provides instructions and examples for executing shell commands, specifically using uvx to run the dewey CLI and executing Python scripts (e.g., btm_safegraph_pull.py) for data processing and export tasks.
[INDIRECT_PROMPT_INJECTION]: The skill serves as a gateway to ingest vast amounts of external structured data (Parquet, CSV) into the agent's environment, which represents a potential attack surface.
Ingestion points: External data is fetched from Dewey Data's download servers and processed using duckdb and deweypy, as seen in examples/btm_safegraph_pull.py and references/duckdb.md.
Boundary markers: There are no specific instructions or delimiters mentioned to prevent the agent from being influenced by potentially malicious instructions embedded within the ingested data content.
Capability inventory: The skill possesses capabilities to perform network requests to deweydata.io and write files to local directories (e.g., ~/projects/batm/).
Sanitization: The skill lacks explicit sanitization or validation logic for the content of the data files before they are processed by the agent.

dewey