312-frameworks-spring-data-jdbc
Pass
Audited by Gen Agent Trust Hub on Sep 26, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill is designed to read and analyze project source code and configuration files to suggest improvements. This creates an indirect prompt injection surface where instructions hidden within the analyzed code could potentially influence agent behavior. \n
- Ingestion points: Existing project files and build configurations read during the initial analysis phase (Workflow Step 1). \n
- Boundary markers: The skill lacks explicit markers or safety instructions to distinguish between the content of the analyzed files and the agent's instructions. \n
- Capability inventory: The agent can execute shell commands via the Maven wrapper (./mvnw) to compile and verify code. \n
- Sanitization: The skill provides guidance on preventing SQL injection in generated code, but does not specify sanitization for the inputs it analyzes. \n- [COMMAND_EXECUTION]: The skill uses shell commands (./mvnw compile, ./mvnw clean verify) to ensure code validity. These are standard development operations used here as safety gates for the refactoring process.
Audit Metadata