blink-rag
Pass
Audited by Gen Agent Trust Hub on Sep 14, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill provides tools for uploading external documents and URLs into a knowledge base for AI-powered search. This creates an indirect prompt injection surface where malicious content within uploaded documents could attempt to influence the agent's behavior during RAG operations.\n
- Ingestion points: Document ingestion occurs via
blink.rag.upload(supporting raw content, URLs, and file uploads) as documented inSKILL.md.\n - Boundary markers: There are no documented boundary markers or instructions provided to the LLM to ignore potentially malicious embedded commands in the retrieved context.\n
- Capability inventory: The skill utilizes SDK methods for document processing (
blink.data.extractFromUrl) and AI search (blink.rag.aiSearch); it does not contain direct command execution or broad file system access.\n - Sanitization: The documentation does not describe any sanitization or validation steps performed on external content before it is indexed or queried.
Audit Metadata