Skills
skills/secondsky/sap-skills/sap-hana-ml/Gen Agent Trust Hub

sap-hana-ml

Pass

Audited by Gen Agent Trust Hub on Mar 15, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
  • [PROMPT_INJECTION]: The skill documents methods for ingesting data from external SAP HANA tables, CSV files, and shapefiles, which establishes an indirect prompt injection surface. If an agent processes untrusted data from these sources without sanitization, it could be manipulated.\n
  • Ingestion points: Multiple entry points are identified in references/DATAFRAME_REFERENCE.md, including ConnectionContext.table(), ConnectionContext.sql(), and import_csv_from().\n
  • Boundary markers: The skill does not currently provide instructions or patterns for using delimiters to isolate untrusted data from the system prompt.\n
  • Capability inventory: The skill enables significant capabilities including executing SQL on remote HANA instances and persisting models/data to the database and local filesystem.\n
  • Sanitization: The documentation includes the hana_ml.dataframe.quotename utility for escaping SQL identifiers.\n- [EXTERNAL_DOWNLOADS]: The skill references standard installation of the hana-ml library and its visualization suite (including matplotlib, plotly, graphviz, and wordcloud) from the Python Package Index (PyPI).
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 15, 2026, 11:24 PM