The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill instructions demonstrate handling incoming webhook data and function parameters (such as CallSid, agent_phone, and conference_name) and interpolating them directly into TwiML string literals. This creates a potential injection surface if the agent or developer uses these patterns to process untrusted user input without applying the recommended VoiceResponse builder.
Ingestion points: External data enters the context via request.form["CallSid"] (Python) and req.body.CallSid (Node.js), as well as through function arguments like agent_phone and new_agent_phone in SKILL.md.
Boundary markers: The skill includes a clear textual security warning stating, "Never interpolate untrusted user input into inline twiml= strings," and suggests using the SDK's VoiceResponse builder for dynamic content.
Capability inventory: The skill uses the Twilio SDK to perform network-based telephony operations, including client.calls.create and participants.update across all scripts.
Sanitization: Code snippets for TwiML attributes use raw string interpolation (f-strings or template literals) without explicit input validation or escaping, relying instead on the accompanying textual warning.
[COMMAND_EXECUTION]: The skill provides numerous patterns for executing telephony commands via the Twilio API, such as initiating calls, placing participants on hold, and muting legs. These are standard functional capabilities for the skill's intended purpose.
[EXTERNAL_DOWNLOADS]: The skill references and downloads the official twilio library for Python and Node.js. These are well-known official packages provided by the skill author (Twilio) and are utilized according to best practices.

twilio-conference-calls