The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests untrusted data from guest interactions, specifically through reservation details and service requests, which could be used to influence agent behavior if the agent processes these inputs as instructions.\n
Ingestion points: Data is ingested in SKILL.md via the submit_guest_request method (guest name, request description) and in references/PROPERTY_MANAGEMENT_SYSTEM.md via create_reservation (guest name, email, phone, special requests).\n
Boundary markers: The skill lacks explicit delimiters or instructions to the agent to treat guest-provided strings as data rather than instructions.\n
Capability inventory: The skill allows access to Read, Write, and Edit tools, providing a surface where an injection could attempt to manipulate the file system.\n
Sanitization: The provided Python logic does not include sanitization or validation of the guest-provided text fields.

hospitality-expert