spring-ai-integration
Pass
Audited by Gen Agent Trust Hub on Oct 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCREDENTIALS_UNSAFE
Full Analysis
- [CREDENTIALS_UNSAFE]: The file
examples/bad-ai-service.javacontains a placeholder credential stringsk-ant-abc123.... This is used exclusively for educational purposes to illustrate insecure coding practices that should be avoided by developers. - [INDIRECT_PROMPT_INJECTION]: The skill implements patterns for processing untrusted user data within LLM prompts, representing a potential attack surface. 1. Ingestion points: Untrusted data enters via method parameters in
AiServiceTemplate.java(e.g.,customerEmail,items) andAiServiceTemplate.javacontroller methods. 2. Boundary markers: The provided prompt templates, such astemplates/analyze-order.st, do not use explicit delimiters or 'ignore embedded instruction' warnings for interpolated variables. 3. Capability inventory: All scripts utilize theChatClientwhich is capable of making network calls to external AI providers (Anthropic, OpenAI). 4. Sanitization: The skill follows best practices by using Spring AI's.param()method for safe interpolation, which helps mitigate basic injection but does not prevent content-based indirect injection.
Audit Metadata