retail-etl-pipeline-medallion
Warn
Audited by Gen Agent Trust Hub on Jun 12, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSCOMMAND_EXECUTIONDATA_EXFILTRATIONPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: The skill instructs users to clone an external repository (github.com/EsraaSolimanMubarak/Retail-ETL-Pipeline) and execute the SQL scripts contained within. This introduces a supply chain risk as the code is from an untrusted third-party source.
- [COMMAND_EXECUTION]: The troubleshooting section includes instructions to use
xp_cmdshellin SQL Server. This feature allows the database engine to execute operating system commands, presenting a high risk of privilege escalation or system compromise if misconfigured or exposed to untrusted input. - [INDIRECT_PROMPT_INJECTION]: The pipeline processes untrusted CSV data from external sources, which constitutes a vulnerability surface for indirect injection.
- Ingestion points: Raw CSV files located in the
/data_source/directory. - Boundary markers: None identified; the skill performs standard bulk ingestion without security boundaries.
- Capability inventory: The skill uses
pyspark,sqlcmd, anddocker-compose, and suggestsxp_cmdshellfor system-level access. - Sanitization: Only basic business logic transformations (e.g., TRIM, UPPER) are applied; there is no evidence of security sanitization for incoming data.
Audit Metadata