roboflow-universe
Pass
Audited by Gen Agent Trust Hub on Oct 29, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to search for and evaluate datasets from Roboflow Universe, which involves processing potentially untrusted third-party metadata.
- Ingestion points: External data enters the agent's context through search results returned by the
universe_searchtool, including dataset names, descriptions, tags, and class lists fromuniverse.roboflow.com. - Boundary markers: The instructions do not specify boundary markers or include explicit warnings to ignore instructions that might be embedded in the retrieved metadata.
- Capability inventory: The skill allows the agent to recommend datasets, provide direct links, and guide the user through forking or downloading data, which could be influenced by malicious metadata.
- Sanitization: The skill does not mention any sanitization or validation steps for the metadata retrieved from the external repository.
Audit Metadata