google-gemini-file-search
Warn
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: MEDIUMEXTERNAL_DOWNLOADSPROMPT_INJECTION
Full Analysis
- Unverifiable Dependencies & Remote Code Execution (MEDIUM): The file
scripts/create-store.tsand thescripts/README.mdfile reference a Node.js package named@google/genai. This package name does not correspond to the official Google Generative AI SDK (@google/generative-ai), making it an unverifiable dependency and a potential typosquatting risk. - Indirect Prompt Injection (LOW): The skill is designed for Retrieval-Augmented Generation (RAG), which involves ingesting external data that could contain malicious instructions. 1. Ingestion points: Document uploads to search stores (planned in
upload-batch.ts). 2. Boundary markers: Absent; no explicit delimiters or 'ignore instructions' warnings are used for retrieved content. 3. Capability inventory:scripts/create-store.tshas the capability to create and manage cloud search stores via the Google AI API. 4. Sanitization: Absent; the skill currently contains no logic to sanitize or validate the content of ingested files.
Audit Metadata