Skills Overview
The Claude Code Toolkit ships with 27 built-in skills that automate common development and data engineering workflows. Each skill is a structured procedure that Claude follows step-by-step, ensuring consistent, high-quality results.
Skills by Category
Git and GitHub
| Skill |
Description |
| PR Workflow |
Lint, test, branch, commit, and open a pull request in one pass |
| TODO to Issues |
File roadmap items and inline TODO/FIXME comments as GitHub issues |
| Review PR |
Check out a PR, review the diff, run tests, and post feedback |
| Protect Main |
Apply branch protection rules via the GitHub API |
| Cleanup Branches |
Delete local and remote branches that have been merged to main |
| Git Rescue |
Recover lost commits, undo bad rebases/resets, and restore deleted branches |
| Secrets Audit |
Scan the working tree, tracked files, and git history for exposed credentials |
| PII Audit |
Scan data files, code, and git history for exposed personal information |
Data Engineering
| Skill |
Description |
| Data Profile |
Generate a data quality report for CSV, Parquet, Excel, or database tables |
| API Explore |
Hit a public API, inspect the response, and generate a typed Python client |
| ETL Scaffold |
Create a standard Extract-Transform-Load pipeline with Pydantic validation |
| Backfill |
Idempotent, resumable, gap-aware historical data loads |
| Migrate |
Versioned, reversible schema migrations for SQLite, Turso, and DuckDB |
| Data Diff |
Compare two datasets and report schema, row-count, and value differences |
| ML Experiment |
Scaffold or audit a model-training script for seeds, leakage, and baselines |
Trading
| Skill |
Description |
| MC Analysis |
Monte Carlo + walk-forward validation, writing a strategy research report |
| Trade Review |
Analyze closed trades and compare live P&L against the backtest |
| Add Event Source |
Onboard a new economic event to the forex trading bot |
Project Setup
| Skill |
Description |
| New Project |
Scaffold a Python project with pyproject.toml, ruff, pytest, and git |
Web Development
| Skill |
Description |
| New Web App |
Scaffold a React + Vite + TypeScript web app with standard tooling |
| Deploy Web App |
Deploy or debug a Vite/React or Next.js app to Vercel, Netlify, or Docker |
| UI Components |
Build typed, accessible React components matching the project's styling approach |
Code Quality
| Skill |
Description |
| Lint Fix |
Auto-detect the project linter and fix all errors without manual intervention |
Dashboards and Visualization
| Skill |
Description |
| Deploy Streamlit |
Deploy or debug a Streamlit app locally, on Streamlit Cloud, or via Docker |
| New Dashboard |
Scaffold a Streamlit or Shiny dashboard over your data source |
| Data Visualization |
Create clear, honest charts following Edward Tufte and Stephen Few principles |
| Infographic Design |
Design a single-page infographic with one clear takeaway and strict visual hierarchy |
How Skills Work
Skills are defined as step-by-step markdown instructions in skills/*/SKILL.md. When you ask Claude to perform one of these tasks, it follows the procedure exactly, using the appropriate CLI tools (gh, git, ruff, pytest, etc.) and applying the good-practice patterns documented in each skill.
Invoking a skill
You do not need to name the skill explicitly. Asking "open a PR for my changes" will trigger the PR Workflow skill. Asking "profile this CSV" will trigger Data Profile. Claude matches your intent to the appropriate skill automatically.
Customization
Each skill can be extended or overridden by editing its SKILL.md file in the skills/ directory. Project-specific conventions defined in CLAUDE.md take precedence over skill defaults when there is a conflict.