sap-cloud-sdk-ai

Pass

Audited by Gen Agent Trust Hub on Mar 10, 2026

Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADSCREDENTIALS_UNSAFE
Full Analysis
  • [EXTERNAL_DOWNLOADS]: The skill references and provides installation instructions for official SAP SDK packages and widely used ecosystem libraries (e.g., LangChain, Zod) from public registries. These sources are considered well-known and trusted technology providers.\n- [CREDENTIALS_UNSAFE]: All credential-related code examples and connection guides use safe placeholders (such as 'clientid':'...', 'client-id', and '<AICORE_INSTANCE>') to demonstrate setup without exposing actual secrets.\n- [PROMPT_INJECTION]: The skill documents features that ingest and process external data, creating an indirect prompt injection surface.\n
  • Ingestion points: Untrusted data enters the agent context through placeholderValues and grounding_input configurations as described in SKILL.md and orchestration-guide.md.\n
  • Boundary markers: The SDK promotes the use of template delimiters like {{?question}} and {{?context}} to separate instructional prompts from external data.\n
  • Capability inventory: The skill provides functionality for tool calling (e.g., weather API fetch in agentic-workflows.md), model orchestration, and management of AI Core resources.\n
  • Sanitization: Proactively provides examples and builders for implementing AzureContentSafetyFilter, LlamaGuardFilter, and DpiMasking to sanitize inputs and outputs.
Audit Metadata
Risk Level
SAFE
Analyzed
Mar 10, 2026, 04:51 AM