inboxmate-inbox-demo
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [COMMAND_EXECUTION]: The skill uses the
curlutility to interact with the vendor's GraphQL API atcrm.psquared.devto create sales opportunities. This is a legitimate functional requirement and uses an Authorization: Bearer token retrieved from a secure environment variable ($PSQUARED_CRM_TOKEN). - [INDIRECT_PROMPT_INJECTION]: The skill ingests data from external prospect websites to generate industry-specific email threads and match company brand tone. 1. Ingestion points: Scraping of prospect website homepages, services, and FAQ pages (Phase 1). 2. Boundary markers: No specific delimiters or "ignore previous instructions" warnings are used when processing the scraped data. 3. Capability inventory: Calls to the
create_inbox_demotool andcurlfor CRM updates. 4. Sanitization: No specific filtering or escaping mechanisms are described for the external content. While this creates a surface for indirect prompt injection, it is the primary intended function of the skill and includes a required verification step (Phase 4).
Audit Metadata