agentphone

Pass

Audited by Gen Agent Trust Hub on Sep 6, 2026

Risk Level: SAFEINDIRECT_PROMPT_INJECTIONCOMMAND_EXECUTION
Full Analysis
  • [INDIRECT_PROMPT_INJECTION]: The skill represents a surface for indirect prompt injection because it ingests and processes untrusted external data.
  • Ingestion points: The agent processes call transcripts (transcript field) and incoming SMS message content (body field) as described in SKILL.md.
  • Boundary markers: There are no explicit instructions or delimiters provided in the prompt examples to distinguish between system instructions and untrusted data from calls/messages.
  • Capability inventory: The skill allows the agent to place outbound phone calls, send SMS messages, and manage account resources using curl commands in SKILL.md.
  • Sanitization: No explicit sanitization or filtering of the transcript or SMS content is mentioned in the instructional logic.
  • [COMMAND_EXECUTION]: The skill uses curl to interact with the AgentPhone API at api.agentphone.to. These commands are used for standard telephony operations such as creating agents, buying numbers, and initiating calls.
Audit Metadata
Risk Level
SAFE
Analyzed
Sep 6, 2026, 07:24 AM
Security Audit — agent-trust-hub — agentphone