shiny-calendarstore
Pass
Audited by Gen Agent Trust Hub on Aug 28, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill allows an agent to ingest and process calendar event data, which acts as an untrusted external source that could contain hidden instructions.
- Ingestion points: Calendar event content (titles, descriptions, locations) is retrieved via
ICalendarStore.GetEventsandICalendarStore.GetEvent(referenced in SKILL.md), as well as through thesearch_eventsandget_eventAI tools. - Boundary markers: The instructions do not specify the use of delimiters or 'ignore' instructions when passing event content to the agent's context.
- Capability inventory: The agent is granted capabilities to create, update, and delete events (
CreateEvent,UpdateEvent,DeleteEventin SKILL.md), which could be targets for an injection attack. - Sanitization: There is no documentation regarding the sanitization or filtering of event content before it is processed by the AI model.
Audit Metadata