Cortex-Nexus: Domain-Specific Emotional Prompt Engineering for Large Language Models

Experimental framework using 912 evaluation cycles across 8 experimental phases (V3C-V6B) with a double-blind LLM-as-a-judge architecture to study emotional prompt injection in large language models. Key findings: (1) Each task domain has a distinct optimal emotional configuration — curiosity for philosophical tasks, concentration for technical tasks, and technical mastery for code generation. (2) Subtle emotional intensity outperforms extreme intensity (the Subtlety Effect). (3) Combining multiple emotional axes produces cognitive interference, not synergy. (4) Self-refinement pipelines amplify emotional effects by 7× in code generation. These findings are formalized as the Task-Emotion Alignment Hypothesis. Statistical validation includes Wilcoxon signed-rank tests with Cohen's d effect sizes (d=0.528 for code generation mastery, p=0.003). Full dataset (912 raw evaluation cycles in JSONL format), experimental engine source code, configuration files, and PeerJ Computer Science LaTeX manuscript included. Repository: https://github.com/SperanzaMax/Cortex-Nexus