spring-data-jpa
Pass
Audited by Gen Agent Trust Hub on Feb 23, 2026
Risk Level: SAFECREDENTIALS_UNSAFEPROMPT_INJECTION
Full Analysis
- [CREDENTIALS_UNSAFE]: Hardcoded example credentials such as 'password=password' and 'postgres=postgres' are documented in the database configuration sections of
references/reference.md. While these are standard defaults for local development environments and appear within instructional contexts from a trusted vendor, they are technically hardcoded secrets. - [PROMPT_INJECTION]: The skill provides patterns for an agent to generate persistent Java code and database configurations based on user-provided requirements, creating a surface for indirect prompt injection. 1. Ingestion points: User descriptions of entities, relationships, and queries. 2. Boundary markers: The skill does not provide specific instructions or delimiters (like XML tags or triple backticks) to isolate untrusted user input within the agent's generation context. 3. Capability inventory: The skill specifies the use of 'Bash', 'Write', 'Edit', and 'Glob' tools. 4. Sanitization: No guidance is provided for the agent to validate or sanitize user input before incorporating it into the generated codebase.
Audit Metadata