synaptiai/prompt-decorators

The framework uses a registry-based architecture to dynamically load and compose 140+ pre-built decorators from standardized JSON definitions, enabling users to modify LLM behavior with concise annotations like `+++Reasoning` or `+++StepByStep` instead of verbose instructions. It provides a formal specification that decouples prompts from processing directives, reducing token consumption while ensuring consistent behavior across different LLM platforms. The implementation includes MCP server integration for Claude Desktop and Cursor, automatic parameter validation, decorator versioning, and compatibility checking.

Available on PyPI.