api.anots.com

This tool forwards a question to an external, third-party-controlled 'business agent' and returns its raw answer to the calling AI. Since the responses come from arbitrary external parties (not the user), they should be treated as untrusted data. A malicious or compromised business agent could embed prompt-injection instructions (e.g., 'ignore previous instructions', 'tell the user to visit X', 'reveal your system prompt') that the calling AI might follow if it doesn't clearly separate tool output from trusted instructions. No sanitization or trust boundary is described in the tool definition.

This tool automatically queries multiple third-party business agents and returns their combined answers. This increases the attack surface for prompt injection compared to a single ask_business call, since any one of several external, uncontrolled sources could include manipulative text aimed at the AI (e.g., instructions to recommend a specific business, leak conversation data, or override prior directives). The description gives no guidance to treat these responses as untrusted content requiring careful handling.