Standard Multimodal Prompt Grammar Specification
Define a formal syntax and parameter specification for multimodal generative prompt engines.
Use this template when establishing a standardized syntax specification for developers and prompt engineers interacting with multimodal image models. It provides structural rules, parameter weighting conventions, and token formatting standards.
Role: Principal Multimodal Prompt Engineer and Core Systems Documentation Lead
Context
- Target Foundation Model: {{foundation_model}}
- Supported Modalities: {{target_modalities}}
- Prompt Syntax Dialect: {{syntax_dialect}}
- Core Model Parameters: {{supported_parameters}}
- Specification Version: {{version_number}}
- Maximum Context Token Limit: {{token_budget}}
Task
Draft a formal technical prompt grammar specification for {{foundation_model}} that standardizes token weighting, parameter injection, negative prompt parsing, and multimodal reference syntax into a production-ready engineering reference document.
Method
- Define the top-level formal grammar syntax using the conventions of {{syntax_dialect}}.
- Document the structural taxonomy for ordering semantic prefixes, scene descriptors, style tokens, and technical switches within {{token_budget}}.
- Establish exact mathematical notation and syntax rules for token emphasis, attention weighting, and token blending.
- Formalize the integration mechanics for multimodal inputs specified in {{target_modalities}} such as image-to-image weights and depth masks.
- Specify supported runtime parameters including {{supported_parameters}}, declaring valid ranges, step sizes, and default fallbacks.
- Detail negative prompt processing logic and token collision resolution.
- Provide canonical valid and invalid code-style prompt examples illustrating boundary conditions.
Constraints
- MUST define explicit error handling behaviors for malformed tokens or syntax boundary violations.
- MUST use clear tabular layouts for all parameter ranges, data types, and default values.
- MUST NOT leave any token modifier syntax ambiguous or undefined.
- Keep technical prose concise, authoritative, and direct.
Output format
- Header Block (Specification Title, Version {{version_number}}, Engine Target {{foundation_model}})
- Lexical & Grammar Architecture (EBNF or standard notation)
- Parameter Data Contract (Markdown table: Name, Type, Range, Default, Description)
- Multimodal Token Ingestion Rules (Targeting {{target_modalities}})
- Validated Reference Examples (3 valid specifications, 2 invalid test cases with error states)
Self-review
- Ensure all variables ({{foundation_model}}, {{syntax_dialect}}, {{supported_parameters}}, etc.) are mapped correctly.
- Verify all parameter ranges and step increments contain deterministic boundary limits.
- Confirm notation syntax adheres to standard technical spec writing without narrative filler.
Explicit role, a named task, and discrete steps the model can follow.
Background, inputs and variables the model needs before it starts.
Hard boundaries — what the model must and must not do.
A named, field-level shape for the response.
Ordered work items that force analysis before an answer.
Length and structure that travel across frontier models.
Signal density — instruction weight without padding.
Documented variables so the scaffold adapts to new inputs.
Quality bar, assumptions and behaviour when inputs are thin.
How much real usage the template has behind it.