The Agent Skills Directory

[COMMAND_EXECUTION]: The instructions provide commands to install and execute the ElevenLabs CLI tool (@elevenlabs/cli) globally via npm to manage agent configurations, authentication, and deployment.
[EXTERNAL_DOWNLOADS]: The documentation for widget embedding includes a script reference to fetch the ElevenLabs conversation widget from the unpkg.com content delivery network.
[INDIRECT_PROMPT_INJECTION]: The skill facilitates the creation of voice agents that ingest untrusted user speech and text. These agents have the capability to trigger webhooks and client-side JavaScript tools, creating a surface for potential instruction injection via user data.
Ingestion points: User transcripts and conversation audio processed in real-time (documented in SKILL.md and references/agent-configuration.md).
Boundary markers: The configuration documentation in references/agent-configuration.md details available guardrails and safety settings to mitigate malicious inputs.
Capability inventory: Agents can execute server-side webhooks (POST/GET), browser-side client tools, and initiate outbound phone calls (documented in references/client-tools.md and references/outbound-calls.md).
Sanitization: The skill supports a sanitize option for dynamic variables to remove sensitive values from tool responses before they are processed by the agent (documented in references/outbound-calls.md).

agents