Tour Booking

Sub-agent for outbound listing-office calls to request and confirm property showing slots using a provided call script and structured payloads. Use when a pa...

This skill implements outbound phone calls and will send job data (client name, address, phone, timezone and other metadata) to ElevenLabs when run in live mode. The bundle and SKILL.md do not declare the two required env vars (ELEVENLABS_API_KEY and ELEVENLABS_AGENT_ID) in the registry metadata — that mismatch reduces transparency. Before enabling live calls: 1) Keep to dry-run mode for testing. 2) Review the included scripts (prepare_call_payload.py and place_outbound_call.py) yourself to confirm what fields are sent. 3) Only provide ELEVENLABS credentials in a secure environment, and consider using credentials with limited scope. 4) Do not set ELEVENLABS_OUTBOUND_URL to an untrusted endpoint (it can redirect where PII is sent). 5) If you need higher assurance, ask the publisher to update the registry metadata to declare required env vars and provide an authoritative homepage/source, or run the skill in an isolated environment first.

Type: OpenClaw Skill Name: tour-booking Version: 0.1.0 The skill is classified as suspicious due to significant vulnerabilities related to input sanitization. The `scripts/prepare_call_payload.py` script directly uses `listing['office_phone']` from user-controlled input as the `to_number` for outbound calls, enabling potential arbitrary phone calls to untrusted destinations. Additionally, the `system_prompt` for the ElevenLabs AI agent is constructed using f-strings with user-controlled inputs (`client_name`, `address`, `preferred_windows_text`), creating a prompt injection vulnerability against the downstream AI agent. While the skill's core function of making outbound calls via ElevenLabs is legitimate, these vulnerabilities could be exploited for malicious purposes if the inputs are not properly sanitized by the calling workflow.