Claude Code skills, ranked and checked

Searches BGPT, a scientific-paper MCP service, for full-text methods, results, sample sizes and quality data beyond abstracts.

Creates and audits editable scientific posters in macro-free PowerPoint, checking physical dimensions, assets, accessibility, provenance and package security.

Creates, edits and analyses Excel workbooks with formulas, formatting, financial models and cleaned spreadsheet data.

Organises PyTorch neural-network projects with LightningModules, data modules, Trainer configuration, callbacks, logging and distributed training.

Applies Aeon, a scikit-learn compatible Python toolkit, to classify, forecast, cluster and analyse temporal data.

Guides exact symbolic mathematics in Python with SymPy, covering algebra, calculus, equation solving, matrices, physics and executable code generation.

Manages Zotero reference libraries with the pyzotero Python client, including bibliographic items, collections, tags, attachments and citation exports.

Generates and edits photos, illustrations, logos and other visual assets through OpenRouter's Image API, with model and output controls.

Loads Hugging Face Transformers models for inference, text generation, tokenisation and Trainer-based fine-tuning across language, vision and audio tasks.

Explains and audits machine-learning predictions with SHAP, selecting explainers, validating feature attributions, and producing local or global visualisations.

Detects effective CPU, memory, disk, scheduler and accelerator limits with Python, then plans conservative resources for local scientific workloads.

Creates professional infographics with Nano Banana Pro, using optional Perplexity research, style presets, accessible palettes and iterative quality review.

Scales pandas, NumPy, text and custom Python workloads beyond memory with Dask's parallel collections, schedulers and clusters.

Guides MATLAB and GNU Octave workflows for numerical code, data, testing, plotting, migration and safe Python interoperability.

Optimises scientific Python workloads for NVIDIA GPUs with CUDA and RAPIDS, choosing suitable libraries, validating results and benchmarking end-to-end speed.

Builds and evaluates right-censored and competing-risk survival models with scikit-survival, using leakage-safe pipelines and censoring-aware metrics.

Migrates agent skills and their supporting files into an existing Warp Factory configuration with adapted roles, handoffs and validation.