sap-cloud-sdk-ai-python
Pass
Audited by Gen Agent Trust Hub on Jul 23, 2026
Risk Level: SAFEEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [EXTERNAL_DOWNLOADS]: Provides instructions to install the official sap-ai-sdk-gen Python package from PyPI. It also suggests the databricks-agent-skills repository as a resource for Databricks environments.
- [CREDENTIALS_UNSAFE]: Documents the configuration of SAP AI Core credentials via environment variables and local JSON configuration files. The examples use safe placeholders and follow standard SDK authentication patterns.
- [PROMPT_INJECTION]: While the skill enables the agent to process external data through LLM calls, creating an indirect injection surface, it provides extensive documentation on mitigating these risks using the SAP Orchestration Service.
- Ingestion points: External data enters via messages in native clients and placeholder_values in the Orchestration Service as seen in SKILL.md and orchestration-guide.md.
- Boundary markers: The skill demonstrates the use of templating delimiters like {{?question}} to separate instructions from data.
- Capability inventory: Capabilities include LLM chat completion, embeddings generation, and vector search via SAP AI Core.
- Sanitization: Explicit guidance is provided for using the filtering module for content safety and the masking module for PII anonymization to sanitize inputs and outputs.
- [SAFE]: No malicious patterns, obfuscation, or unauthorized data exfiltration attempts were detected. The skill aligns with official SAP documentation for AI Core integration.
Audit Metadata