harness-advanced-elicitation

LLM output refinement applies method-driven elicitation techniques like socratic questioning and red-teaming to existing drafts, forcing deeper reflection and iterative enhancement. It selects the best five methods based on context and applies them in strict order to replace the section with an improved version pending user confirmation.

Red-teaming and first-principles analysis are applied through an interactive refinement loop that loads elicitation methods, presents five best-fit options for your draft, and executes chosen techniques in strict order to produce an enhanced text section requiring user confirmation before applying changes.

Iterative elicitation targets specific sections of recent LLM output by loading methods from a configuration file, allowing you to pick, reshuffle, or list all available critique methods, and replacing the original section content with an enhanced version after explicit user confirmation.

Socratic questioning and pre-mortems are set up by instructing your agent to enhance the current section using advanced elicitation after generating a draft, then selecting from numbered methods presented based on the context of your content for structured application.

Structured elicitation specifically targets existing LLM-generated drafts rather than new generations, applying selected critique methods to replace section content with enhanced versions and requiring explicit user confirmation before finalizing any changes to the original output.