awesome-ai-agent-skills: Skill for Claude Code

Academic & Scientific Research Skill

Sourced from danielmiessler/fabric research patterns (extract_paper_wisdom, create_design_document, summarize_paper) and academic publication standards (IEEE / ACM / Springer).

---

1. Literature Review & SOTA Synthesis

• Systematic Search & Taxonomization: Categorize existing literature by methodology, dataset assumptions, performance limits, and open research gaps.
• Paper Wisdom Extraction (extract_paper_wisdom):
  • Extract: Primary Hypothesis, Core Novelty/Contribution, Theoretical Claims, Empirical Benchmark Results, Baseline Comparisons, Limitations & Future Work.
• State-of-the-Art (SOTA) Comparison Matrix: Construct quantitative comparison tables summarizing accuracy/F1/latency/FLOPs across competing papers.

---

2. Research Hypothesis & Methodology Design

• Problem Formulation: Mathematically define inputs $\mathcal{X}$, target outputs $\mathcal{Y}$, mapping functions $f_\theta(x)$, and objective loss functions $\mathcal{L}(\theta)$.
• Research Hypotheses ($H_1, H_2$): State testable, falsifiable research hypotheses before conducting experiments.
• Methodology Specification:
  • Architectural block diagrams (Mermaid / TikZ).
  • Algorithmic pseudocode (LaTeX algorithm2e or algorithmicx).
  • Mathematical derivations with explicit index and dimension notations.

---

3. Experimental Benchmarking & Reproducibility

• Rigorous Evaluation Metrics: Precision, Recall, F1-score, Mean Squared Error (MSE), BLEU/ROUGE, Inference Latency (ms), Memory Footprint (MB), Statistical Significance ($p$-value $< 0.05$).
• Ablation Studies: Systematic isolation of individual system components to measure individual contribution to performance.
• Reproducibility Guarantee:
  • Seed initialization (random_seed = 42).
  • Exact hyperparameter logging (learning rate, batch size, weight decay, epoch count).
  • Version-controlled dataset splits.