The Agent Skills Directory

[PROMPT_INJECTION]: The skill possesses an indirect prompt injection attack surface because it processes untrusted data from OData URLs and SAP object names.
Ingestion points: The skill ingests untrusted data through the url parameter in odata_perf and object name parameters in SAPContext, SAPRead, and SAPDiagnose calls.
Boundary markers: There are no boundary markers or explicit instructions provided to the agent to ignore or delimit potentially malicious instructions embedded within the processed OData URLs or metadata fields.
Capability inventory: The skill utilizes high-privilege vendor tools including SAPQuery for arbitrary SQL execution and SAPDiagnose for starting and stopping system-level traces.
Sanitization: The instructions do not define any sanitization, validation, or escaping logic for the external content before it is passed to the execution tools.
[SAFE]: The skill utilizes vendor-specific tools (SAPQuery, SAPDiagnose, etc.) provided by the author 'arc-mcp'. These represent standard functionality for SAP debugging and do not constitute unauthorized command execution.

debug-slow-sql