rag-pgvector
Pass
Audited by Gen Agent Trust Hub on Aug 3, 2026
Risk Level: SAFE
Full Analysis
- [SAFE]: The skill provides technical architectural instructions for setting up a RAG pipeline, which is a standard development workflow for AI agents.
- [SAFE]: It explicitly promotes secure secret management by advising that API keys (such as OpenAI keys) should only be stored as server-side secrets (Edge Function secrets) and never requested or provided in chat sessions.
- [SAFE]: The skill includes security-best-practice guidance for data isolation, recommending the use of Supabase Row Level Security (RLS) and metadata filtering to prevent unauthorized data access between users.
- [SAFE]: It provides specific prompt-engineering constraints intended to mitigate indirect prompt injection and hallucinations by instructing the model to respond only using the retrieved context and to admit when information is missing.
Audit Metadata