The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill retrieves data from external providers including Google Sheets, Google Drive, and Amazon S3 through tools like create_table_from_csv. These are recognized well-known services used for the skill's primary data ingestion purpose.
[COMMAND_EXECUTION]: The query_data tool executes SQL SELECT queries. It incorporates security restrictions that block data modification keywords (INSERT, UPDATE, DELETE) and file-reading functions (read_csv_auto, glob) to prevent unauthorized access or system manipulation.
[DATA_EXFILTRATION]: The workflow requires CSV files to be publicly accessible for import. This creates a potential risk of data exposure if users host sensitive information on public URLs for the purpose of importing it, though the intended operation moves data into the user's managed environment.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection (Category 8) due to the ingestion of untrusted data from external URLs. Ingestion points: Data enters the agent's context via create_table_from_csv and other update tools described in SKILL.md. Boundary markers: No explicit instructions or delimiters are provided to the agent to treat the imported CSV or spreadsheet content as data rather than instructions. Capability inventory: The skill allows for SQL querying and database modification through specific insert and update tools. Sanitization: While technical SQL keywords are filtered in the query tool, there is no validation or sanitization of the natural language content within the imported files which could potentially influence agent behavior.

gentic-data