agentphone
Pass
Audited by Gen Agent Trust Hub on Sep 23, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill processes untrusted input from phone calls and SMS messages, which could potentially be used by an external party to influence agent behavior through indirect prompt injection. 1. Ingestion points: Inbound call transcripts and SMS messages processed via webhook events (SKILL.md). 2. Boundary markers: The instruction examples do not include explicit prompt delimiters or instructions for the AI to ignore embedded commands. 3. Capability inventory: The skill enables the agent to initiate outbound calls, send SMS, and manage account resources such as agents and phone numbers (SKILL.md). 4. Sanitization: While the documentation advises cleaning LLM output to remove non-speech artifacts, it does not provide specific sanitization or validation logic for inbound telephony data.
Audit Metadata