nshskill 0.1.0

Ship Claude Code skills from Python packages with zero boilerplate.

nshskill

Ship Claude Code skills from Python packages with zero boilerplate.

Why

Claude Code skills are markdown files that live in .claude/skills/ and teach Claude how to use your library. If you maintain a Python package, you probably want users to install your skill with a single command — but wiring up the file copying, frontmatter parsing, and CLI yourself is tedious and repetitive. nshskill handles all of that so you can focus on writing the skill content.

Install

pip install nshskill
# or
uv add nshskill

Requires Python 3.10+.

Quick start

Option A: Let Claude do it

Install the init-skill skill, then invoke it in Claude Code:

nshskill skill install          # install to current project
nshskill skill install --global # install at user level

Then in Claude Code, run /init-skill and Claude will analyze your codebase and set up the full nshskill integration — _skill/ directory, SKILL.md, CLI wiring, and dependency.

Option B: Manual setup

1. Create a skill directory

Add a _skill/ directory inside your package with a SKILL.md file:

src/myproject/
├── __init__.py
├── ...
└── _skill/
    ├── SKILL.md
    └── references/         # optional extra files
        └── api-cheatsheet.md

2. Write SKILL.md with frontmatter

The file must start with YAML frontmatter containing at least a name field:

---
name: using-myproject
description: Short description of when Claude should use this skill.
---

# myproject

Instructions for Claude on how to use your library go here.

3. Add a CLI entry point

The simplest approach — a standalone CLI with just the skill command:

# src/myproject/cli.py
from pathlib import Path
from nshskill import Skill, create_skill_cli

skill = Skill.from_dir(Path(__file__).resolve().parent / "_skill")
main = create_skill_cli("myproject", skill)

# pyproject.toml
[project.scripts]
myproject = "myproject.cli:main"

Users can then run:

myproject skill install          # install to .claude/skills/using-myproject/
myproject skill install --global # install to ~/.claude/skills/using-myproject/
myproject skill uninstall        # remove it

API

Skill

from nshskill import Skill

A dataclass with two fields:

Field	Type	Description
name	str	Skill name (used as the directory name under .claude/skills/)
skill_dir	Path	Path to the source directory containing SKILL.md

Skill.from_dir(skill_dir: Path) -> Skill

Parse a Skill from a directory. Reads the name from the YAML frontmatter in SKILL.md. Raises FileNotFoundError if the directory or SKILL.md is missing, and ValueError if the frontmatter is missing or lacks a name field.

skill = Skill.from_dir(Path(__file__).resolve().parent / "_skill")

skill.install(target=None, *, global_=False) -> Path

Copy the skill directory contents into <target>/.claude/skills/<name>/. Returns the destination path.

• If global_=True, installs to ~/.claude/skills/<name>/ (ignores target).
• If target is None, defaults to the current working directory.
• Overwrites any existing installation at the destination.
• Skips __init__.py and __pycache__ when copying.

skill.uninstall(target=None, *, global_=False) -> bool

Remove an installed skill directory. Returns True if something was removed, False if the skill wasn't installed at that location.

add_skill_commands

from nshskill import add_skill_commands

def add_skill_commands(subparsers, skill: Skill) -> None

Add skill install and skill uninstall subcommands to an existing argparse subparsers action. Use this when your package already has a CLI and you want to add skill management alongside other commands:

import argparse
from pathlib import Path
from nshskill import Skill, add_skill_commands, dispatch_skill

def main():
    parser = argparse.ArgumentParser(prog="myproject")
    subs = parser.add_subparsers(dest="command")

    # Your other commands...
    subs.add_parser("run", help="Run the thing")

    # Add skill install/uninstall
    skill = Skill.from_dir(Path(__file__).resolve().parent / "_skill")
    add_skill_commands(subs, skill)

    args = parser.parse_args()

    if args.command == "skill":
        dispatch_skill(args)
    elif args.command == "run":
        ...

dispatch_skill

from nshskill.cli import dispatch_skill

def dispatch_skill(args: argparse.Namespace) -> None

Route a parsed skill subcommand to the appropriate install/uninstall method. Call this after parse_args() when args.command == "skill".

create_skill_cli

from nshskill import create_skill_cli

def create_skill_cli(prog: str, skill: Skill) -> Callable[[], None]

Return a ready-to-use main() function for packages that only need skill commands in their CLI. The returned function parses sys.argv and dispatches automatically:

# src/myproject/cli.py
from pathlib import Path
from nshskill import Skill, create_skill_cli

main = create_skill_cli("myproject", Skill.from_dir(Path(__file__).resolve().parent / "_skill"))

Skill directory layout

A skill directory can contain any files you want to ship to users. The only requirement is a SKILL.md at the root:

_skill/
├── SKILL.md              # required — frontmatter + skill instructions
└── references/           # optional — supplementary docs
    ├── api.md
    └── examples.md

To include existing docs from your repo without duplicating them, use symlinks:

ln -s ../../docs/api.md src/myproject/_skill/references/api.md

When installed, the entire directory tree is copied to .claude/skills/<name>/, preserving structure. __init__.py and __pycache__ are excluded automatically.

SKILL.md format

---
name: using-myproject
description: When to activate this skill.
---

Your skill content in markdown. This is what Claude reads
to learn how to use your library.

The frontmatter must contain:

• name (required): Used as the subdirectory name under .claude/skills/.
• description (optional): Helps Claude decide when the skill is relevant.

License

MIT