twilio-conversation-intelligence
Pass
Audited by Gen Agent Trust Hub on Sep 15, 2026
Risk Level: SAFEINDIRECT_PROMPT_INJECTIONEXTERNAL_DOWNLOADSCOMMAND_EXECUTION
Full Analysis
- [INDIRECT_PROMPT_INJECTION]: The skill manages conversation analysis from external channels such as SMS and Voice, which are untrusted inputs. Malicious actors could use these channels to send adversarial instructions intended for the AI processing the conversation. The documentation identifies this risk and recommends using boundary instructions in LLM prompts to ignore embedded commands.
- [EXTERNAL_DOWNLOADS]: Integration patterns in the documentation demonstrate network requests to official Twilio API endpoints (intelligence.twilio.com, conversations.twilio.com, insights.twilio.com) for configuration management and results retrieval.
- [COMMAND_EXECUTION]: The skill defines rules for webhook actions that trigger external POST requests based on conversation events. While this is a primary feature for workflow automation, it enables the automated transmission of analysis results to external endpoints.
Audit Metadata