The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The documentation references official ElevenLabs packages and libraries across multiple registries, including npm (@elevenlabs/cli, @elevenlabs/elevenlabs-js, @elevenlabs/client, @elevenlabs/react, @elevenlabs/react-native), PyPI (elevenlabs), and the unpkg CDN (@elevenlabs/convai-widget-embed). These are standard vendor-managed resources.
[REMOTE_CODE_EXECUTION]: The installation guide includes a shell installer command (curl ... | sh) for the official ElevenLabs CLI. The source targets the authorized vendor repository at github.com/elevenlabs/cli, which is a recognized and trusted source for this platform.
[CREDENTIALS_UNSAFE]: The skill follows security best practices by instructing users to store the ELEVENLABS_API_KEY in environment variables rather than hardcoding it. All code examples use standard placeholders like your-api-key or {{API_KEY}} for configuration templates.
[PROMPT_INJECTION]: The skill provides detailed guidance on configuring safety guardrails (platform_settings.guardrails) specifically designed to detect and block prompt injection attempts and instruction overrides.
[INDIRECT_PROMPT_INJECTION]: The documentation describes how the agent ingests untrusted data from knowledge base documents and webhook tool outputs. It provides mitigations such as markdown boundary markers for prompts and a sanitize option for dynamic variables to remove sensitive data from tool responses before they are processed by the LLM.
[SAFE]: The analyzed content consists entirely of legitimate developer documentation, API references, and implementation examples for a voice AI platform, with no evidence of malicious intent or hidden behaviors.

agents