inboxmate-demo
Pass
Audited by Gen Agent Trust Hub on Oct 7, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted data from external websites and uses it to construct the system prompt and knowledge base for generated chatbots.
- Ingestion points: Scrapes prospect website content (homepage, about, products, FAQ, contact) using
WebFetchin Phase 1b. - Boundary markers: Absent; the scraped text is directly interpolated into the system prompt template in Phase 2c and knowledge item content in Step 3.3.
- Capability inventory: Possesses significant capabilities including creating/updating agents (
create_agent,update_prompt) and adding data to vector stores (add_to_bucket). - Sanitization: Absent; the skill does not specify any filtering or validation for the content scraped from external domains before using it to define agent behavior.
- [EXTERNAL_DOWNLOADS]: Performs network requests to external domains for business logic.
- Evidence: Calls the OpenBrand API at
openbrand.shto extract color and logo data, and the CRM API atcrm.psquared.devfor opportunity tracking. - [COMMAND_EXECUTION]: Uses shell commands to interact with external services.
- Evidence: Employs
curlwithin its workflow to send and receive data from the OpenBrand and CRM endpoints.
Audit Metadata