api-vector-db-weaviate
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill facilitates Retrieval Augmented Generation (RAG), which involves fetching data from an external vector database and interpolating it into prompts for an LLM. This creates a surface for indirect prompt injection if the stored data contains malicious instructions.
- Ingestion points: Data from Weaviate collections is ingested via the
generate.*submodule (e.g.,articles.generate.nearTextinexamples/search.md). - Boundary markers: The provided examples lack boundary markers or explicit instructions to the LLM to ignore instructions embedded within retrieved properties. For example, Pattern 8 in
examples/search.mduses `"Summarize this article in one tweet: {title} - {body}"`.
- Capability inventory: The skill enables fetching and processing external data from a database that could be controlled by an attacker.
- Sanitization: No sanitization or validation of the retrieved properties is described before they are interpolated into the prompt.
Audit Metadata