ai-persistence/build-drizzle-adapter

Pass

Audited by Gen Agent Trust Hub on Sep 27, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill defines logic for storing chat messages and session data in a database. This involves processing data originating from external requests.
  • Ingestion points: Ingests messages, threadId, and runId from the POST request in the route handler implementation.
  • Boundary markers: The implementation uses Drizzle ORM's query builder, which provides parameterized query safety for all database interactions.
  • Capability inventory: The skill performs database writes (insert/update/delete) across the chat_threads, chat_runs, chat_interrupts, and chat_metadata tables in src/lib/chat-persistence.ts.
  • Sanitization: The skill explicitly instructs developers to authorize thread access at the route level by deriving the user from a session instead of trusting client-supplied IDs, mitigating potential access control issues.
  • [COMMAND_EXECUTION]: The instructions guide the developer to use standard CLI tools for database management.
  • Evidence: The text recommends running npx drizzle-kit generate and wrangler d1 migrations apply to sync the database schema. These are well-known tools from established providers (Drizzle and Cloudflare) and are necessary for the skill's primary purpose of database integration.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 27, 2026, 04:07 PM