The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses SAPWrite and SAPActivate to create and activate CDS views within the connected SAP system. This is the primary and intended function of the skill, utilized for legitimate development tasks.
[EXTERNAL_DOWNLOADS]: The skill utilizes a search tool (referencing mcp-sap-docs) to retrieve current SAP documentation and best practices for analytical annotations. This information is used as a reference to ground the code generation process.
[DATA_EXPOSURE]: The skill reads structural metadata from the SAP system using SAPRead (e.g., table field lists, class definitions, and CDS element definitions). This data is used internally to generate the new model and is not exfiltrated to external domains.
[INDIRECT_PROMPT_INJECTION]: The skill has an attack surface for indirect prompt injection as it processes untrusted metadata from the SAP system (table/view definitions) and interpolates this into generated ABAP DDL code.
Ingestion points: SAP metadata retrieved via SAPRead(type="TABL"), SAPRead(type="DDLS"), and user-provided identifiers in SKILL.md.
Boundary markers: Not explicitly present in the ABAP source templates.
Capability inventory: The skill possesses SAPWrite and SAPActivate capabilities, allowing it to modify the SAP system state.
Sanitization: No specific escaping or validation of the retrieved SAP metadata is documented before it is used in code generation.
Conclusion: While the surface exists, the risk is inherent to development tools of this nature and is considered a standard operational risk within the development environment.

generate-analytics-star-schema