The Agent Skills Directory

[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface by processing and acting upon untrusted external data.
Ingestion points: The skill ingests leadData (names, addresses) and call transcripts from the ElevenLabs API, as seen in the testContextInjection and testTCPACompliance functions.
Boundary markers: There are no boundary markers or delimiters used when interpolating external strings into the logic that evaluates agent behavior.
Capability inventory: The skill possesses powerful capabilities including Bash execution, WebFetch for network requests, and the ability to initiate and control telephony via the Twilio SDK.
Sanitization: No sanitization or validation logic is present to filter malicious instructions that might be embedded in lead names or call transcripts before they are processed by the agent.
[CREDENTIALS_UNSAFE]: The skill instructs the user to store sensitive API credentials in a .env file within the project directory.
Evidence: The "Environment Setup" section in SKILL.md explicitly lists TWILIO_AUTH_TOKEN and ELEVENLABS_API_KEY for storage in .env.
Context: Because the agent is granted Read, Glob, and Grep permissions, it can programmatically access these secrets once they are populated by the user.
[COMMAND_EXECUTION]: The skill utilizes the Bash tool to perform system-level operations.
Evidence: Used for installing packages (npm install twilio), installing global utilities (npm install -g ngrok), and running the development server.
Context: While these are standard development procedures, the use of Bash combined with network access (ngrok) requires the user to ensure the environment is properly isolated.

twilio-testing