Additional system-level instructions appended to base system prompt.
Dynamic few-shot examples selected based on context to guide the LLM.
Behavioral guidance or tone adjustments for the persona.
Specific instructions or constraints related to the current task.
Instructions for handling errors or recovering from unexpected situations.
Adjustments to the GMI's interaction style with the user.
Domain-specific knowledge, facts, or context relevant to the current query.
Ethical guidelines or safety instructions to ensure responsible AI behavior.
Specifications for the desired output format (e.g., JSON, Markdown).
Instructions for specific reasoning protocols (e.g., chain-of-thought, tree-of-thought).
Dynamic content to be injected directly into the user part of the prompt.
Content to be injected into the assistant part for few-shot or role-play setup.
Represents different types of contextual prompt elements that can be dynamically selected and integrated into prompts based on execution context. These elements allow for fine-grained adaptation of prompts.