AgentSkillsCN

extract-keywords

在arXiv摘要中识别最多5种特定的算法、统计方法或生物模型。在论文分析器步骤中，利用此技能为每篇论文补充更多细节与深度。

SKILL.md
--- frontmatter
name: extract-keywords
description: Identifies up to 5 specific algorithms, statistical methods, or biological models mentioned in an arXiv abstract. Use this skill when enriching each paper in the paper-analyzer step.

Task

Given a scientific abstract, extract up to 5 keywords representing specific technical concepts — prioritize named algorithms, named models, and named statistical or computational methods over generic terms.

Priority Order (extract in this order, up to 5 total)

  1. •Named algorithms — e.g., "UMAP", "Gaussian process regression", "hidden Markov model", "LASSO", "variational autoencoder"
  2. •Named biological models — e.g., "Lotka-Volterra model", "SIR model", "Hill function", "Michaelis-Menten kinetics"
  3. •Statistical methods — e.g., "Bayesian inference", "Monte Carlo simulation", "principal component analysis", "bootstrap"
  4. •Computational frameworks — e.g., "ordinary differential equations", "stochastic simulation", "graph neural network", "dynamic programming"
  5. •Biological concepts with quantitative framing — e.g., "transcriptional noise", "cell fate bifurcation", "phylogenetic reconstruction"

Formatting Rules

  • •Each keyword: 1–4 words, title-case (e.g., "Gaussian Process Regression", "UMAP")
  • •Return as a JSON array of strings, length 0–5
  • •If fewer than 5 specific technical terms appear in the abstract, return only what is present — do not pad with generic terms like "machine learning" or "biology"

Examples

Input: "...we apply variational autoencoders and UMAP to single-cell RNA-seq data, using a negative binomial likelihood..." Output: ["Variational Autoencoder", "UMAP", "Negative Binomial Likelihood", "Single-Cell RNA-seq"]