New Skills

design-taste-frontend is an anti-slop frontend skill for landing pages, portfolios, editorial pages, and redesigns. It helps an agent read the brief, infer the right visual direction, and ship interfaces that feel intentional instead of templated. Best for Frontend Development when design taste, hierarchy, and brand fit matter.

design-taste-frontend-v1 is the original taste-skill for high-agency frontend work, preserved for exact backward compatibility. It helps generate React/Next.js-oriented UI code with clearer taste decisions, dependency checks, and practical guardrails. Use this design-taste-frontend-v1 skill when you need the older behavior instead of the v2 experimental rewrite.

create-skill-test scaffolds eval.yaml test files for agent skills in dotnet/skills. Use it to create skill tests, define scenarios, fixtures, assertions, and rubrics, and reduce overfitting in evaluation design. It is not for running existing tests, debugging validator errors, or authoring SKILL.md files.

create-skill is a scaffold generator for new agent skills in the dotnet/skills style. Use it to create a valid skill folder, generate SKILL.md with frontmatter, and follow repository conventions for Skill Scaffolding. It is best for new skills, not editing existing ones.

create-custom-agent helps you create VS Code custom agent files (.agent.md) for specialized AI personas with tools, instructions, and handoffs. Use it to scaffold new agents, set tool limits, and define agent-to-agent workflows for Skill Authoring.

write is an editing skill for rewriting Chinese or English prose so it sounds natural, clear, and less AI-like while preserving the original meaning. It fits drafts, docs, release notes, launch copy, social posts, and write for Technical Writing cleanup. Use it when you already have text and want tighter, audience-aware prose, not fresh ideation from zero.

think is a decision-support skill for turning rough ideas into approved, decision-complete plans before coding. Use it for feature design, architecture choices, tradeoff analysis, and should-we-do-this questions where the goal is judgment, not implementation. It fits think for Decision Support, think guide, and think usage needs in repo-first workflows.

The read skill fetches URLs and PDFs as clean Markdown for reading, quoting, citation, and downstream work. It is built for read usage on paywalled pages, JS-heavy sites, X/Twitter, GitHub files, Chinese platforms, and Workflow Automation flows that need reliable source text before analysis. Use the read guide when you want source capture, not commentary.

Learn is a research skill that turns unfamiliar topics, source bundles, and collected notes into a coherent, publish-ready output. It helps with deep dives, source compilation, explanation, and structured synthesis for web research and other multi-source tasks. Best when you need one strong reference from many inputs, not a quick lookup.

hunt is a debugging-first skill that forces root-cause thinking before any fix is applied. Use it for errors, crashes, regressions, failing tests, stale cache issues, screenshot bugs, and “it used to work” failures. It helps you find a testable hypothesis, gather evidence, and avoid guesswork. Not for code review or new features.

health runs a budget-aware Agent Health audit for Codex, Claude Code, Pi, agent instructions, hooks/MCP, verifier surfaces, and AI maintainability. Use the health skill to check why an agent ignores instructions, misses validation, or drifts into hard-to-maintain behavior. It is especially useful for Security Audit workflows, but not for debugging code or reviewing PRs.

The design skill helps turn vague UI requests into production-grade visual output for pages, components, dashboards, and screenshot-driven polish. Use it when the interface looks ugly, unclear, inconsistent, or visually wrong, and when you need design for UI Design rather than backend logic or data pipelines. It includes guidance for install, usage, guardrails, and better aesthetic decisions.

The check skill reviews code diffs, PRs, issue queues, release readiness, commits, pushes, publishing, and project audits. Use it when you need a disciplined check for Code Review before merge or release, with safety gates for dirty and untracked worktrees. It is not for exploring ideas, debugging root causes, or prose review.

The treatment-plans skill helps generate concise, clinician-focused medical treatment plans in LaTeX/PDF-ready form. It supports general medicine, rehabilitation, mental health, chronic disease, perioperative care, and pain management with SMART goals, evidence-based interventions, minimal citations, and compliance-aware formatting. Ideal for treatment-plans for Technical Writing and structured care documentation.

The transformers skill helps you use Hugging Face Transformers for model loading, inference, tokenization, and fine-tuning. It is a practical transformers guide for Machine Learning tasks across text, vision, audio, and multimodal workflows, with clear paths for quick baselines and custom training.

torchdrug is a PyTorch-native toolkit for molecular and protein machine learning. Use the torchdrug skill to choose tasks, datasets, and modular models for graph neural networks, protein modeling, knowledge graph reasoning, molecular generation, and retrosynthesis. It is best for custom model development and reproducible configs, not just canned demos.

torch-geometric skill guide for PyTorch Geometric graph neural networks. Use it for torch-geometric install help, torch-geometric usage, graph classification, node classification, link prediction, heterogeneous graphs, custom MessagePassing layers, and scaling GNNs for Machine Learning workflows.

Use the sympy skill for exact symbolic math in Python, including algebra, calculus, matrices, physics formulas, number theory, geometry, and code generation. It helps you keep expressions exact, choose the right SymPy modules, and avoid float-heavy mistakes. Best for users who need a practical sympy guide for symbolic workflows and sympy for Data Analysis.

The statsmodels skill helps you use statsmodels for data analysis in Python when you need statistical models, inference, and diagnostics. It fits OLS, GLM, discrete outcomes, time series, and mixed models, with coefficient tables, p-values, confidence intervals, and assumption checks. Use this statsmodels guide for econometrics, forecasting, and defensible reporting.

The statistical-analysis skill helps you choose, run, and report defensible tests for Data Analysis, including assumptions, effect sizes, power, and APA-style results. Use it for academic research, experiments, and observational studies when test selection and clear reporting matter more than coding a specific model.

stable-baselines3 skill guide for Machine Learning workflows: train RL agents, wire Gymnasium environments, and choose PPO, SAC, DQN, TD3, DDPG, or A2C with less guesswork. Best for standard single-agent reinforcement learning, quick prototyping, and practical stable-baselines3 usage.

simpy is a Python framework for process-based discrete-event simulation. This simpy skill helps model queues, resources, and time-based events for manufacturing, service operations, logistics, networking, and simpy for Data Analysis when you need wait time, utilization, throughput, or bottleneck insights.

shap skill for model interpretability and explainable AI. Use it to understand predictions, compute feature attributions, choose SHAP plots, and debug model behavior for Data Analysis across tree, linear, deep learning, and black-box models.

Seaborn is a seaborn skill for Python statistical visualization with pandas-friendly inputs and strong defaults. Use it for quick exploration of distributions, relationships, categorical comparisons, box plots, violin plots, pair plots, and heatmaps. Built on matplotlib for static, publication-ready charts.