Glossary¶
This glossary defines key terms and concepts used in the Prompt Decorators framework.
A¶
Annotation¶
The syntax used to apply a decorator to a prompt, consisting of the decorator prefix (+++), the decorator name, and optional parameters.
Example: +++StepByStep(numbered=true)
Apply¶
The process of transforming a prompt by executing the decorator's transform function.
C¶
Category¶
A grouping of decorators with similar functionality or purpose, such as "Reasoning", "Format", or "Style".
Claude Desktop¶
A desktop application for interacting with Anthropic's Claude AI model. Prompt Decorators integrates with Claude Desktop through the MCP protocol.
Composite Decorator¶
A decorator that internally uses multiple other decorators to create a more complex transformation.
Conflict¶
A situation where two or more decorators have incompatible behaviors or modify the same aspect of a prompt in contradictory ways.
D¶
Decorator¶
A reusable component that modifies a prompt to enhance how an LLM processes and responds to it.
Decorator Definition¶
A specification of a decorator, including its name, description, category, parameters, and transform function.
Decorator Instance¶
A specific instance of a decorator with particular parameter values.
Decorator Pattern¶
A design pattern where objects (decorators) wrap around other objects (prompts) to modify their behavior without changing their structure.
Decorator Registry¶
A catalog of available decorators that can be looked up by name.
Dynamic Decorator¶
A decorator that is defined and loaded at runtime rather than being generated as code.
Dynamic Implementation¶
The approach used in Prompt Decorators v0.3.0+ where decorators are loaded directly from definitions at runtime, without code generation.
I¶
Inline Syntax¶
The syntax for including decorators directly in prompt text using the +++ prefix.
Example:
Integration¶
A connection between the Prompt Decorators framework and another system, such as an LLM provider or client application.
J¶
JavaScript¶
The language used for writing transform functions in decorator definitions.
L¶
Large Language Model (LLM)¶
An AI model trained on vast amounts of text data that can generate human-like text based on prompts.
M¶
MCP (Model Context Protocol)¶
A protocol for integrating tools with LLM clients, allowing them to extend the capabilities of the LLM.
MCP Integration¶
The component of Prompt Decorators that implements the Model Context Protocol, enabling integration with MCP-compatible clients like Claude Desktop.
MCP Server¶
A server that implements the Model Context Protocol and exposes Prompt Decorators functionality as MCP tools.
P¶
Parameter¶
A configurable option for a decorator that affects its behavior.
Example: In StepByStep(numbered=true), numbered is a parameter with the value true.
Parameter Type¶
The data type of a parameter, which can be string, number, boolean, or enum.
Precedence¶
The rule determining which decorator takes effect when multiple decorators conflict. In Prompt Decorators, later decorators take precedence over earlier ones.
Prompt¶
The input text sent to an LLM to generate a response.
Prompt Engineering¶
The practice of crafting prompts to effectively communicate with and get desired outputs from LLMs.
R¶
Registry¶
See Decorator Registry.
S¶
Sandboxed Environment¶
A restricted execution environment where JavaScript transform functions run, preventing access to external resources for security.
Stacking¶
The practice of applying multiple decorators to a single prompt, either inline or programmatically.
T¶
Token¶
The basic unit of text that LLMs process. Tokens are typically word fragments, with a word consisting of one or more tokens.
Token Limit¶
The maximum number of tokens that an LLM can process in a single request, including both the prompt and the response.
Transform Function¶
A JavaScript function that defines how a decorator modifies a prompt. It takes the input text and parameter values and returns the transformed text.
Transformation¶
The process of modifying a prompt by applying one or more decorators.
V¶
Validator¶
A tool for checking decorator definitions against the framework's specifications and identifying potential issues.
Common Decorator Categories¶
Audience Decorators¶
Decorators that adapt content for specific audience expertise levels.
Examples: Audience, ELI5, Technical
Domain Decorators¶
Decorators that focus on specific knowledge domains or fields.
Examples: Scientific, Legal, Medical
Format Decorators¶
Decorators that control the structure and format of the output.
Examples: OutputFormat, Bullet, TableFormat
Length Decorators¶
Decorators that control the verbosity or conciseness of responses.
Examples: Concise, Detailed, Summary
Persona Decorators¶
Decorators that adopt specific roles, personalities, or expertise.
Examples: Persona, AsExpert, Professional
Reasoning Decorators¶
Decorators that enhance logical thinking and problem-solving.
Examples: Reasoning, StepByStep, TreeOfThought
Style Decorators¶
Decorators that modify tone, voice, and writing style.
Examples: Tone, Academic, Creative
Common Decorator Parameters¶
depth¶
Controls the level of detail or thoroughness in a response.
Possible values: basic, moderate, comprehensive
format¶
Specifies the output format for structured responses.
Possible values: markdown, json, html, csv, etc.
level¶
Indicates the expertise level for audience adaptation.
Possible values: beginner, intermediate, expert
numbered¶
Determines whether steps or items should be numbered.
Possible values: true, false
role¶
Specifies the role or persona to adopt.
Possible values: Various roles like teacher, scientist, lawyer, etc.
style¶
Defines the writing or communication style.
Possible values: formal, casual, technical, creative, etc.