MiroMindAI
Code Duplication & Maintainability Problems in Prompt Generators
The current implementation of generate_mcp_system_prompt, generate_no_mcp_system_prompt, generate_agent_specific_system_prompt, and generate_agent_summarize_prompt contains extensive code duplication. Large blocks of boilerplate text are repeated across functions, making it difficult to maintain and extend. Adding a new agent type or modifying instructions requires editing multiple places in the code.
Problems Identified:
Repetition of Long Strings:
Many identical or near-identical text blocks (e.g., cautionary instructions, task strategies) are repeated across functions.
Hard to update consistently.
Scalability Issue:
Adding new agent types requires modifying multiple if/elif branches instead of a single mapping/dictionary.
High risk of introducing inconsistencies.
Poor Separation of Concerns:
Long instructional strings are hardcoded in Python functions rather than stored in reusable constants, templates, or even external files.
Readability & Maintainability:
Lack of type hints and docstrings reduces clarity.
JSON schemas for tools are dumped inline, which can reduce readability when large.
Suggested Improvements:
Refactor into reusable templates: Use base templates and format with variables instead of repeating blocks.
Use dictionary mappings instead of long if/elif: Map agent types (main, agent-browsing, etc.) to their specific template text.
Extract long instructional text into constants or Markdown template files: Keep Python code lean and focused on logic, not large strings.
Pretty-print JSON schemas with json.dumps(..., indent=2): For readability.
Add type hints & docstrings: Improve IDE support and maintainability.
Improve error handling: When raising ValueError, list available agent types to help debugging.
