knowledge-base-injection
Pass
Audited by Gen Agent Trust Hub on Sep 20, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill implements a retrieval-augmented generation (RAG) pattern that ingests external knowledge patterns and campaign data into the agent's prompt context.
- Ingestion points: Data enters the agent's context via an external pattern loading mechanism (referenced as
gads-knowledge.js) and user-supplied campaign data. - Boundary markers: The implementation utilizes
<domain_knowledge>and</domain_knowledge>tags to delimit injected content, which provides structural separation but does not prevent malicious content within the data from influencing the model. - Capability inventory: Capabilities are limited to context assembly and LLM interaction; no dangerous capabilities such as arbitrary shell execution or unauthorized network writes were identified.
- Sanitization: Basic tokenization and punctuation stripping are performed, but the system lacks formal sanitization or escaping for the data injected into the prompt.
Audit Metadata