initiate-inboxmate-knowledge
Pass
Audited by Gen Agent Trust Hub on Apr 24, 2026
Risk Level: SAFEPROMPT_INJECTION
Full Analysis
- [SAFE]: The skill is purely informational, providing an architectural overview of the InboxMate ecosystem.
- [SAFE]: References to environment variables such as $PSQUARED_CRM_TOKEN and $NUXT_MCP_DEMO_TOKEN follow standard security practices for defining configuration requirements without hardcoding secrets.
- [SAFE]: All identified API endpoints and domains (e.g., crm.psquared.dev, app.psquared.dev) correspond to the author's official infrastructure.
- [PROMPT_INJECTION]: The skill documents an architecture that retrieves and processes data from external systems (CRM, databases), which inherent to the system's design, represents an indirect prompt injection surface.
- Ingestion points: Data is retrieved from Twenty CRM, Supabase, and the Notification Service APIs.
- Boundary markers: The documentation does not specify the use of delimiters or 'ignore' instructions for the agent when handling these external data sources.
- Capability inventory: The architecture references the use of SQL execution, GraphQL queries, and REST API calls across the pipeline.
- Sanitization: No specific validation or sanitization protocols for external data are described in this architectural overview.
Audit Metadata