The Agent Skills Directory

[CREDENTIALS_UNSAFE]: The skill documents the storage of API keys and profiles in the local file system at ~/.twilio-cli/config.json. It also references the use of sensitive environment variables such as TWILIO_AUTH_TOKEN and SENDGRID_API_KEY for authentication.
[EXTERNAL_DOWNLOADS]: The instructions include commands to download and install the Twilio CLI from several sources, including npm, Homebrew, and a third-party Scoop bucket (https://github.com/twilio/scoop-twilio-cli). It also suggests the installation of the ngrok tunneling service.
[COMMAND_EXECUTION]: The skill provides extensive examples of shell commands for resource provisioning, phone number management, and serverless project initialization, which involve active manipulation of the local environment and Twilio account resources.
[DATA_EXFILTRATION]: The skill includes a functional example of piping the output of ps aux directly to twilio email:send. This pattern demonstrates how sensitive system process information can be transmitted to an external email address.
[INDIRECT_PROMPT_INJECTION]: As a reference guide for an AI agent, this skill creates a surface where user-supplied input (e.g., message bodies, email recipients, webhook URLs) is interpolated into shell commands. Without strict boundary markers, this could lead to command injection or unintended resource manipulation.
[DYNAMIC_EXECUTION]: The skill documents the twilio plugins:install command, which allows the dynamic installation of third-party or vendor-provided packages. This mechanism involves downloading and executing external code at runtime to extend the CLI's functionality.

twilio-cli-reference