Security audit
Security checks for vulnerabilities and agentic risk
This prompt-only rental-listing skill is not malicious, but it asks users to paste live property access details and WiFi credentials into external AI prompts without secure-handling guidance.
Review before installing. Use this only with placeholders such as {{ACCESS_CODE}} and {{WIFI_PASSWORD}}, then substitute real values inside a trusted booking or property-management platform. Do not paste live lock codes, reusable WiFi passwords, unit access procedures, or emergency contact details into a general AI chat. Rotate any real access details that were entered into prompts or copied from generated examples.
prompts/04-automated-guest-messages.md:19Live Property Access Codes and Wi-Fi Credentials Are Submitted to Third-Party AI Services
The file discloses unit number, floor, parking location, and smart-lock delivery workflow, which are physical security details that could facilitate unauthorized entry or targeting of the property. In skill context, this is more dangerous because the artifact is an example/template rather than a protected guest communication, so the information is exposed without access control or stated need.
The skill explicitly asks users to provide operationally sensitive information for automated guest messages, including lockbox code or smart lock details, WiFi credentials, parking instructions, and checkout instructions, but gives no guidance on secure handling, minimization, or avoiding unnecessary disclosure. If this data is entered into an AI tool, stored in logs, or exposed through prompt history or third-party processing, it could enable unauthorized property access, network access, or social engineering against hosts and guests.
The markdown includes sensitive physical and network access information with no warning, classification, or handling constraints, making accidental reuse or publication more likely. The surrounding skill context does not justify public disclosure; as a sample artifact, it should model safe treatment of secrets rather than expose them plainly.
The document publishes a live-looking WiFi SSID/password and related stay details in a broadly reusable artifact, which exposes operational access information beyond the intended guest audience. Even if this is an example skill, embedding realistic credentials trains unsafe handling of secrets and could enable unauthorized network access if the details are real or reused elsewhere.
The automated message templates normalize sending sensitive information such as WiFi credentials and smart-lock codes inside reusable content artifacts, increasing the chance that downstream agents or users will disclose secrets insecurely. While the lock code is templated, the pattern still encourages unsafe secret handling and could result in accidental exposure through logs, previews, or misrouted messages.
This markdown file handles potentially sensitive access information by asking for the WiFi SSID and password and later directing the model to include the actual password in the FAQ. The skill description does not provide any warning or caution about exposing credentials, rotating guest networks, or limiting use to non-sensitive guest WiFi.
The FAQ explicitly instructs the model to include the actual WiFi network name and password from the inputs in the response. This is a natural-language disclosure instruction for sensitive access information, even though it is framed as a normal hospitality task rather than using explicit data-leak terminology.
The prompt explicitly instructs users to include operationally sensitive information such as lockbox or smart-lock access codes and WiFi credentials in automated guest messages. Even though this is a legitimate hospitality workflow, embedding secrets directly into reusable templates without any warning, scoping guidance, or advice on timing increases the risk of accidental disclosure, message misrouting, reuse after checkout, or exposure to unauthorized recipients.