ares-agent-toolkit

High-fidelity academic document parsing (MinerU) + scientific literature retrieval (SciVerse), with one-command setup. Use when the user wants to parse/convert a PDF, image, Word, PPT, or Excel into Markdown/JSON — especially papers with formulas, tables, or scanned/multi-column layout ("解析PDF", "PDF转markdown", "提取论文内容", "公式转LaTeX", "mineru", "pdf2md"); to search scientific papers / literature ("文献检索", "sciverse", "找论文", "semantic search papers"); or to bootstrap this toolchain on a new device ("配置文档解析环境", "新设备装 mineru/sciverse"). For generic PDF ops (merge, split, rotate, fill forms, plain text extract) prefer the built-in `pdf` skill.

Apply the "learning mechanics" framework — the emerging physics-style theory of deep learning training dynamics — when writing, debugging, scaling, or tuning neural-network code. Use for decisions about learning rate / batch size / width / depth scaling, μP (Maximal Update Parameterization) and hyperparameter transfer, edge-of-stability and progressive sharpening, lazy vs. rich (feature-learning) regimes and initialization scale, neural scaling laws, gradient-flow conservation laws / symmetries, neural collapse, the neural feature ansatz, and for designing scientific experiments on training. Trigger phrases: "learning mechanics", "why is training unstable", "how should I scale the learning rate with width/batch", "muP / mup / hyperparameter transfer", "edge of stability", "lazy vs rich", "feature learning regime", "scaling laws", "tune hyperparameters on a small model". Source: Simon et al., "There Will Be a Scientific Theory of Deep Learning" (2026).