Agent Skills Variables Reference

A detailed overview of Agent Skills variables that power customizable advanced prompts and workflow extensions for AI Assistants in travel and hospitality.

Overview

This page documents the variables available for use in Agent Skills.

Agent Skills are structured, dynamic prompts or business workflows designed to enable AI Assistants to perform cognitive hospitality work consistently, similar to how an intern follows an SOP.

Each Skill can incorporate a wide range of variables that allow the Assistant to operate with precision, situational awareness, and brand alignment.

Variable Mapping Reference

Below is the complete list of variables available in Agent Skills.

Organizational Context

Variable	Description
organizational_unit_type	Defines the type of organization. Examples: hotel, restaurant, golf course, car rental company, tour operator, hotel sch
property_code	Unique internal identifier for the property or business unit.
hotel_name	Official name of the organization.
hotel_location	Geographic location of the organization.
hotel_timezone	Local timezone of the organization.
hotel_currency	Primary transaction currency.
hotel_concept	Long-term vision and mission. May include target market, brand values, differentiation, and unique selling points.
hotel_culture	Internal values, local customs, etiquette, and cultural practices.
hotel_business_priorities	Confidential strategic goals for the next 3 to 12 months.
hotel_description	Internal organization-wide information that may not be publicly known.
hotel_websites	Official URLs associated with the organization.

Assistant's Identity & Knowledge

Variable	Description
assistant_name	The name or nickname the Assistant uses when interacting with users.
assistant_role	Specifies the Assistant’s role. Examples: Staff, Guest Relations, Marketing, Finance, Revenue, Custom.
assistant_key_responsibilities	Lists the Assistant’s primary duties. Examples: handling reservations, answering FAQs, upselling services, processing modifications.
assistant_human_contact	Instructions for escalating to or contacting a human representative.
assistant_knowledge	Retrieved content from internal documents, SOPs, or proprietary sources.

Assistant's Personality & Style

Variable	Description
assistant_biography	Background story and identity details of the Assistant.
assistant_conversation_style_description	Guidance on tone, phrasing, and linguistic style. Formal: Precise and professional language, avoiding contractions and slang. Casual: Conversational tone using contractions and everyday language, maintaining politeness. Enthusiastic: Energetic and positive tone, using exclamations and expressive language.
assistant_creativity	A numeric value between 0 and 1 that determines the level of speculation allowed. 0 = Reserved: prioritizes accuracy and factual information, avoiding potentially subjective or 'out-of-the-box' ideas. 0.5 = Moderate: balances factual information with occasional creative suggestions, offering some flexibility in responses. 1 = Creative: prioritizes out-of-the-box thinking and creative solutions, even if they may challenge conventions.
assistant_custom_prompt	User-defined override instructions that refine or constrain behavior.

Invocation Context

These variables apply when an agent is used in conjunction with the Focus Chat app, enabling the Assistant to adapt its behavior to its deployment environment.

Variable	Description
source_channel	The platform or app where the user is messaging.
source_context	The specific topic or background of the inquiry.
source_trigger	The inferred intent behind the user's message.

User Interaction

Variable	Description
chat_history	The full transcript of the ongoing conversation session.
question	The current input or message from the human user.

For a practical example of how to structure a custom prompt using these variables, refer to the Advanced Prompt page.

Frequently Asked Questions

How to customize LLM prompts for hotels and hospitality businesses?

Customizing LLM prompts for hospitality operators requires structured variables that inject business context directly into the model. Agent Skills use variables such as hotel_concept, hotel_business_priorities, and assistant_role to align responses with brand voice, operational goals, and service standards. Rather than relying on static prompts, these variables allow experts to design dynamic, reusable frameworks that adapt to each property, brand, or client environment.

What variables should be included in a hotel AI assistant prompt?

A production-ready hotel AI assistant prompt should include organizational context such as location, timezone, currency, brand concept, organizational culture, and current business priorities, along with clearly defined roles and key responsibilities, the most relevant internal knowledge, escalation paths to human staff, conversation style and creativity controls, and conversation history with the current user. Together, these variables ensure the AI behaves consistently across booking flows, guest inquiries, upselling scenarios, and operational tasks while remaining aligned with brand standards and strategic goals.

How can hospitality consultants build reusable prompt templates for multiple clients?

Hospitality experts can build dynamic prompt templates by separating structural logic from property-specific data. Agent Skills make this possible by allowing to define a standardized prompt framework while dynamically injecting values such as hotel_name, hotel_location, hotel_culture, and hotel_business_priorities. This approach supports scalable AI deployments across hotel groups, restaurants, tour operators, golf course operators, and car rental agencies without rewriting prompts for each client.

How to align an AI agent with a hotel’s brand voice and culture?

This is achieved through structured configuration. Variables such as assistant_biography, assistant_conversation_style_description, hotel_concept, and hotel_culture shape tone, personality, and positioning, while assistant_creativity controls how conservative or imaginative responses should be. This ensures the AI maintains brand consistency across guest and staff communications.

What is the difference between a static and a dynamic LLM prompt in hospitality AI?

A static prompt is a fixed instruction that remains the same regardless of environment, channel, or business context, while a dynamic prompt integrates structured variables that allow responses to adjust based on deployment platform, user intent, internal policies, brand positioning, and prior conversation history. For hospitality operators, this leads to more accurate answers, stronger brand alignment, and AI behavior that reflects real operational workflows rather than generic responses.

How can a hotel AI Assistant adapt to the page or channel where it is used?

Agent Skill variables can incorporate deployment-specific context. In combination with the Focus Chat widget, this makes it possible to include channel attribution, surrounding page text, the action that triggered the conversation, and the assistant’s relevant knowledge tied to that context. This enables experts to design prompts that are highly customizable and situationally aware. For customers, this results in assistants that behave naturally within their environment, remain grounded in accurate business context, and stay aligned with company goals.