Lets Share some Claude Skills

This directory is a git-controled directory for skills developed for Claude.

What's here

share-claude-skills
├── .claude
│   └── skills
│       └── git-commit-checklist
│           └── SKILL.md
├── skills
│   ├── build-q2-pluggins
│   │   ├── references
│   │   │   ├── action-registration.md
│   │   │   ├── external-tools.md
│   │   │   ├── package-structure.md
│   │   │   ├── testing.md
│   │   │   └── type-system.md
│   │   ├── README.md
│   │   └── SKILL.md
│   ├── snakemake
│   │   ├── references
│   │   │   ├── cluster-execution.md
│   │   │   ├── config-and-samples.md
│   │   │   ├── project-structure.md
│   │   │   ├── rule-patterns.md
│   │   │   └── snakemake-style-guide.rst.txt
│   │   ├── README.md
│   │   └── SKILL.md
│   └── when-am-i
│       ├── references
│       │   ├── calendar-context.md
│       │   └── milestones.md
│       ├── deadlines.md
│       └── SKILL.md
├── CLAUDE.md
├── .gitignore
└── README.md

Use These Skills

To use these as your person Claude skills, link your ~/.claude/skills directory to this one. Alternatively, link individual skill folders to use select skills from this set while keeping your own local set not in this repo.

Option 1: To use the whole folder wholesale:

ln -s $PWD/skills ~/.claude/skills

Replace $PWD with the path to this repo or run command from this folder.

Option 2: To use individual skills:

ln -s $PWD/skills/some-skill-1 ~/.claude/skills/some-skill-1
ln -s $PWD/skill/some-skill-2 ~/.claude/skills/some-skill-2
...

Option 1 is recommended for a repository that you maintain individually to for your personal collection. Option 2 is recommended for a shared repository where other people may add skills that you only selectively want to give to Claude.

Where Claude Looks for Skills

Location	Path	Applies to	How many?
Enterprise	managed settings	All users in your organization	One
Personal	~/.claude/skills/<skill-name>/SKILL.md	All your projects	One
Project	.claude/skills/<skill-name>/SKILL.md	This project only	Many
Plugin	<plugin>/skills/<skill-name>/SKILL.md	Where plugin is enabled	Many

When skills share the same name across levels, higher-priority locations win: enterprise > personal > project. Plugin skills are namespaced as plugin-name:skill-name so they can never conflict with other levels.

Personal skills live in exactly one place: ~/.claude/skills/. There is only one such folder per user.

Project skills are not limited to the repo root. Claude discovers .claude/skills/ at the directory where you launched claude, and also in any subdirectory you're actively working in. In a monorepo, packages/frontend/.claude/skills/ is loaded automatically when you're editing files under packages/frontend/. There can be many .claude/skills/ locations within a single project.

Plugin skills are namespaced and isolated per plugin. Each installed plugin contributes its own skills/ directory. You can have any number of plugins active at once, each adding their own skills without conflicting.

Notes

Potential gotchas

provided by Claude.ai

• The --add-dir flag grants file access rather than configuration discovery, but skills are an exception: .claude/skills/ within an added directory is loaded automatically and picked up by live change detection, so you can edit those skills during a session without restarting. The question is whether live change detection follows symlinks — it likely does on most systems, but worth testing.

• If you symlink the entire ~/.claude/skills directory (option 1), make sure the repo root structure matches what Claude Code expects — i.e., each skill is a subdirectory with a SKILL.md inside.

• On macOS, symlinks are followed normally by most tools. On some Linux setups with inotify-based watchers, symlink traversal can occasionally behave unexpectedly, but it's rarely a problem in practice.

Front Matter for skills

https://code.claude.com/docs/en/skills.md — "Frontmatter reference" section

This is still a changing ecosystem. Do not treat this as a comprehensive list.

Field	Type	Description
name	string	REQUIRED Skill identifier (should match directory name)
description	string	REQUIRED Trigger conditions (model-invoked) or /help text (user-invoked)
version	string	for humans Semantic version, e.g. 1.0.0
argument-hint	string	Shown to user, e.g. <required> [optional]
allowed-tools	list	Pre-approved tools, reduces permission prompts
model	string	Override model: haiku, sonnet, opus
disable-model-invocation	bool	true = user-only (for skills with side effects)
user-invocable	bool	false = Claude-only (background knowledge)
context	string	fork = run in isolated subagent
agent	string	Which agent type when forked (e.g. Explore)

Other front matter fields are ignored. When no allowed-tools are given, the skills has the same tool permissions has the main conversation. By default, the skill is invokable by the main conversation (disable-model-invocation: false) when Claude thinks the current task matches the description and invokable by the user as a /command (user-invocable: true).