ai-persistence/build-prisma-adapter
Pass
Audited by Gen Agent Trust Hub on Sep 27, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: No security issues were detected. The skill provides a legitimate template for integrating Prisma as a persistence layer, utilizing well-known libraries and recommending best practices such as manual database migrations and server-side authorization.
- [INDIRECT_PROMPT_INJECTION]: The skill establishes a standard data ingestion surface by allowing the agent to load historical chat transcripts and metadata from a database.
- Ingestion points: Chat history, runs, and metadata are retrieved from the database in the
ChatPersistenceimplementation withinsrc/lib/chat-persistence.ts. - Boundary markers: The implementation does not include specialized delimiters for the LLM, which is expected for a database persistence adapter.
- Capability inventory: The skill defines read and write access to specific database models (
ChatThread,ChatRun,ChatInterrupt,ChatMetadata) via the app'sPrismaClient. - Sanitization: The adapter stores and retrieves content as-is, relying on the application's route-level authorization and the LLM's own safety guardrails.
Audit Metadata