The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill recommends installing official ElevenLabs packages, such as @elevenlabs/cli and the Python elevenlabs library. It also mentions a web widget script hosted on unpkg.com, which is a standard method for integrating the service into websites.
[CREDENTIALS_UNSAFE]: The documentation correctly identifies the need for an ElevenLabs API key and instructs users to store it in environment variables (ELEVENLABS_API_KEY). It also describes the CLI's standard behavior of storing credentials locally for authentication.
[INDIRECT_PROMPT_INJECTION]: The skill describes an agent platform with an attack surface for indirect prompt injection. Ingestion points: User transcripts, Knowledge Base documents, and Webhook responses. Boundary markers: The platform provides built-in guardrails (platform_settings.guardrails) to mitigate injection attempts. Capability inventory: The agents can perform network calls (Webhooks), browser actions (Client Tools), and telephony tasks (Transfers). Sanitization: A 'sanitize' option is available for dynamic variables to handle sensitive data in tool responses.
[DYNAMIC_EXECUTION]: The platform supports extending agent capabilities through client-side JavaScript tools and server-side webhooks, which are standard features for the described service.
[COMMAND_EXECUTION]: The documentation provides examples of using the ElevenLabs CLI to manage projects and configurations, which is a common developer workflow for this platform.

agents