The Agent Skills Directory

[PROMPT_INJECTION]: The skill exposes a potential indirect prompt injection surface through its voice message handling. Ingestion points: The message parameter in the initiate_call, continue_call, and speak_to_user actions in SKILL.md. Boundary markers: No delimiters or instructions are provided to the agent to treat this data as untrusted. Capability inventory: Ability to broadcast synthesized speech to phone numbers via third-party APIs. Sanitization: No content filtering or validation is described.
[COMMAND_EXECUTION]: The skill documents usage of the openclaw voicecall CLI command for initiating calls and inspecting status, which are standard operations for the platform.
[DATA_EXPOSURE]: The skill includes documentation on where to configure sensitive provider credentials (e.g., Twilio auth tokens) but does not contain any hardcoded secrets.

voice-call