The Agent Skills Directory

[PROMPT_INJECTION] (MEDIUM): The skill is vulnerable to Indirect Prompt Injection because it processes external content (SMS bodies, IVR inputs) and possesses the capability to perform actions with side effects (sending messages, making calls). * Ingestion points: The body parameter in TwilioSMS.send_sms and request.form in the Flask IVR routes. * Boundary markers: None are implemented to isolate untrusted data from the agent's instructions. * Capability inventory: Includes client.messages.create for SMS and response.dial for voice redirection. * Sanitization: Validation is limited to phone number formatting; no sanitization is performed on message content.
[DATA_EXFILTRATION] (LOW): The skill implements outbound communication via the Twilio API. This provides a mechanism for an agent to transmit data to external recipients, which could be exploited for exfiltration if the agent is manipulated.
[EXTERNAL_DOWNLOADS] (LOW): The code depends on the twilio and flask Python packages. These are trusted, industry-standard libraries, but they are external dependencies that should be version-controlled in production environments.

twilio-communications