umbraco-search-provider
Pass
Audited by Gen Agent Trust Hub on Feb 17, 2026
Risk Level: SAFEPROMPT_INJECTIONEXTERNAL_DOWNLOADS
Full Analysis
- Indirect Prompt Injection (LOW): The skill workflow involves fetching external data and using it to generate implementation code, creating a Category 8 vulnerability surface.
- Ingestion points: External URLs via the
WebFetchtool (e.g., https://docs.umbraco.com/umbraco-cms/customizing/extending-overview/extension-types). - Boundary markers: Absent; there are no instructions to the agent to treat fetched content as untrusted or to use delimiters.
- Capability inventory: The skill allows the use of
WriteandEdittools to generate and modify files based on the untrusted input. - Sanitization: Absent; the agent is directed to follow fetched documentation directly for implementation.
- Unverifiable Dependencies & Remote Code Execution (LOW): The skill uses
WebFetchto download content fromdocs.umbraco.com. Although this is a legitimate documentation site, it is not on the specific trusted source list. The pattern of fetching external content to drive logic is noted as a low-severity risk in this context.
Audit Metadata