Agent Skills: Reusable Plugins for AI Coding Agents
I use skills.sh to manage reusable agent skills. Skills are packages of instructions, scripts, and resources that AI coding agents load dynamically — think plugins for your agent. They follow the open Agent Skills spec and work across 40+ agents (Claude Code, Cursor, Copilot, Gemini CLI, etc.).
Managing Skills with npx skills⌗
The npx skills CLI is how you find, install, update, and remove skills.
Finding skills⌗
# Find skills interactively or by keyword
npx skills find
npx skills find "bayesian"
# List available skills in a repo before installing
npx skills add <owner/repo> --list
Installing skills⌗
# Install all skills from a repo (symlinked by default, easy to update)
npx skills add <owner/repo>
# Install globally (into ~/)
npx skills add <owner/repo> -g
# Install a specific skill only
npx skills add <owner/repo> -s skill-name
# Target a specific agent
npx skills add <owner/repo> -a claude-code
Updating and removing skills⌗
# Check for updates and apply them
npx skills check
npx skills update
# List and remove installed skills
npx skills list
npx skills remove <skill-name>
Scaffolding a new skill⌗
npx skills init my-skill
Lockfile-Based Installation⌗
If you want a reproducible skill set across machines, you can use a lockfile approach. Installing skills into a project populates a skills-lock.json that you can commit. On another machine, restore the same set with:
npx skills experimental_install
Skills I’ve Published⌗
These live in brojonat/llmsrules and can be installed with:
npx skills add brojonat/llmsrules
| Skill | Description |
|---|---|
fastapi-service |
FastAPI with JWT auth, structlog, Prometheus metrics |
python-cli |
Click CLI with composable subcommand groups |
go-service |
Go microservice with stdlib HTTP, sqlc, urfave/cli, slog |
scikit-learn |
ML pipelines, cross-validation, hyperparameter tuning, MLflow |
k8s-deployment |
Docker multi-stage builds, kustomize overlays, Makefile automation |
ducklake |
DuckLake lakehouse: snapshots, time travel, schema evolution |
openai-agents |
OpenAI Agents SDK: tools, handoffs, context, webhooks (Python + Go) |
pyproject-config |
pyproject.toml with setuptools, ruff, pytest |
ibis-data |
Database-agnostic data access with Ibis |
parquet-analysis |
Parquet file analysis with Ibis and DuckDB |
temporal-go |
Temporal workflows, activities, workers, signals, sagas in Go |
temporal-python |
Temporal workflows, activities, workers, signals, sagas in Python |
Third-Party Skills Worth Checking Out⌗
| Repository | Focus | Highlights |
|---|---|---|
| anthropics/skills | Anthropic official | frontend-design, pdf, docx, xlsx, pptx, skill-creator |
| obra/superpowers | Dev methodology | systematic-debugging, test-driven-development, dispatching-parallel-agents, writing-plans |
| pymc-labs/agent-skills | Probabilistic programming | pymc-modeling (Bayesian stats, PyMC v5+, ArviZ, BART), marimo-notebooks |
| marimo-team/skills | marimo notebooks | marimo-notebook, jupyter-to-marimo, streamlit-to-marimo, anywidget, implement-paper |
| vercel-labs/agent-skills | React / Next.js | vercel-react-best-practices, web-design-guidelines, deploy-to-vercel |
| supabase/agent-skills | PostgreSQL | supabase-postgres-best-practices |
| temporalio/skill-temporal-developer | Temporal | Official Temporal developer skill |
| planetscale/database-skills | MySQL / PlanetScale | PlanetScale database development skills |
Read other posts