dotnet-semantic-kernel
Pass
Audited by Gen Agent Trust Hub on Sep 17, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONDYNAMIC_EXECUTIONEXTERNAL_DOWNLOADS
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides patterns for building LLM applications that process untrusted external data, creating a potential surface for indirect prompt injection.
- Ingestion points: Untrusted data is ingested into the kernel through variables such as
userInput,articleText, andquestionacross various code snippets inSKILL.md. - Boundary markers: The documentation illustrates the use of structured prompt templates (YAML, Handlebars) with role-based markers (e.g.,
<message role="system">), which helps define boundaries. - Capability inventory: The code examples include capabilities for network communication (AI services, vector stores) and local file reading (
File.ReadAllText). - Sanitization: While the skill promotes structured prompt management, it does not demonstrate specific input sanitization or filtering logic within the code snippets.
- [DYNAMIC_EXECUTION]: The skill demonstrates the creation of executable functions from YAML configurations at runtime using
kernel.CreateFunctionFromPromptYaml. This is a core feature of the library but involves dynamic assembly of logic from string templates. - [EXTERNAL_DOWNLOADS]: The skill references official Microsoft NuGet packages and links to the official Semantic Kernel repository for samples and documentation. These references to trusted organizational resources are documented as safe.
Audit Metadata