A portable kit of Agent Skills for Claude and other AI agents.
Reusable skills that help delivery work move faster — drafting solution docs, shaping presentations, running compliance checks, and anything else you do often enough to be worth teaching an agent once. Each skill is a small folder of instructions that teaches an agent to do one thing well: drop it into your tool of choice and the agent picks it up automatically when it's relevant, or you invoke it by name. This repo is a place to collect, version, and share those skills.
New to skills? A skill is just a
SKILL.mdfile (plus any optional supporting files) with a name and a description. The agent reads the description to decide when the skill applies, then loads the full instructions only when it needs them. They follow the open Agent Skills standard, so the same folder works across Claude Code, claude.ai, the Claude API, and other tools that support it.
| Skill | What it does | Triggers when… |
|---|---|---|
conventional-commits |
Writes clear git commit messages that follow the Conventional Commits spec. | you're committing code, want a commit message, or mention changelogs / semver. |
conventional-commitsships as a working example of the SKILL.md format — skills can automate almost any repeatable task, from documents and decks to reviews and checklists, not just code. Keep it, tweak it, or delete it, then add your own with the help oftemplates/SKILL.mdand CONTRIBUTING.md.
Pick your tool below. The two fastest paths first, then tool-specific detail.
Clone once, and every skill is a folder you can copy or symlink wherever your tool expects it:
git clone https://github.com/stokapps/FieldKit.gitPrefer a one-liner? The community skills CLI
installs a skill straight from this repo into most tools:
# Installs into the target you choose (claude-code, cursor, …)
npx skills add stokapps/FieldKit/skills/conventional-commits
npx skillsis a third-party convenience tool, not an Anthropic product. The manual steps below always work and are worth understanding once.
Claude Code loads skills from two filesystem locations — no upload, no restart:
| Scope | Path | Available in |
|---|---|---|
| Personal | ~/.claude/skills/<skill-name>/SKILL.md |
all your projects |
| Project | <repo>/.claude/skills/<skill-name>/SKILL.md |
that repo (commit it to share with your team) |
Copy a skill into one of them:
# Personal — available everywhere
cp -r FieldKit/skills/conventional-commits ~/.claude/skills/
# …or project-scoped, checked in for your team
mkdir -p .claude/skills
cp -r FieldKit/skills/conventional-commits .claude/skills/Prefer to keep a single source of truth and pull updates with git pull? Symlink instead
of copying:
ln -s "$(pwd)/FieldKit/skills/conventional-commits" ~/.claude/skills/conventional-commitsAdding a skill under an existing ~/.claude/skills/ or .claude/skills/ takes effect in
the current session. (Creating one of those top-level folders for the very first time needs
a Claude Code restart so it can start watching it.)
Use it: ask something that matches the description ("write a commit message for my
staged changes") and Claude loads it automatically, or invoke it directly with
/conventional-commits. Confirm it's registered with "What skills are available?".
The Claude Code extension reads the same directories as the terminal, so installation is identical:
- Personal skills →
~/.claude/skills/(works in every workspace). - Project skills →
.claude/skills/inside the workspace folder. Commit it and everyone who opens the repo in the extension gets the skill.
# From your project root, inside VS Code's integrated terminal:
mkdir -p .claude/skills
cp -r /path/to/FieldKit/skills/conventional-commits .claude/skills/Then invoke skills from the Claude panel exactly as in the terminal (/conventional-commits,
or let Claude trigger them). No extension-specific setup is required.
There's also a community "Agent Skills" marketplace extension for VS Code (
formulahendry.agent-skills) that lets you browse and install skills from a UI. It's third-party — the folder-copy above is the official path.
Custom skills are available on Pro, Max, Team, and Enterprise plans with file creation / code execution enabled.
-
Zip the skill. claude.ai wants an archive that contains the skill folder:
scripts/package-skill.sh skills/conventional-commits # -> dist/conventional-commits.zip -
In claude.ai, open Settings → Capabilities → Skills (labeled Features on some plans) and upload the
.zip. -
Start a chat and ask naturally — Claude uses the skill when your request matches its description.
Skills uploaded here are per-user: each teammate uploads their own, and they don't sync to the API or to Claude Code. See the Help Center: Using Skills in Claude.
The desktop app uses the skills enabled on your claude.ai account. Upload the skill on claude.ai (steps above), then manage which skills are active from Customize in the Desktop sidebar. Once enabled, Claude in the desktop app applies the skill automatically when it's relevant.
Skills run inside the code execution tool's
container. Upload a custom skill through the /v1/skills endpoints, then reference its
skill_id in the request container:
# 1. Upload (returns a skill_id). Requires the Skills beta header.
curl https://api.anthropic.com/v1/skills \
-H "x-api-key: $ANTHROPIC_API_KEY" \
-H "anthropic-beta: skills-2025-10-02" \
-F "display_title=conventional-commits" \
-F "file=@dist/conventional-commits.zip"Then in the Messages request, enable the code execution tool, add the
anthropic-beta: skills-2025-10-02 header, and pass the skill in the container.skills
list by skill_id. Uploaded skills are workspace-wide (all workspace members can use
them). Note the API sandbox has no network access and can't install packages at
runtime. Full walkthrough:
Using Agent Skills with the API.
Agent Skills is an open standard, so any tool that supports it consumes the exact same skill folder. Two ways in:
- Drop it in the tool's skills directory. Many agents follow the
.claude/skills/convention; check your tool's docs for the path it watches, and copy the skill folder there. - Use the cross-tool installer:
npx skills add stokapps/FieldKit/skills/<name> -a <tool>(targets includeclaude-code,cursor,codex, and more).
See agentskills.io for the spec and the current list of supported tools.
Everything an agent needs is in the folder. Only SKILL.md is required:
conventional-commits/
├── SKILL.md # frontmatter (name + description) and the instructions
└── references/
└── commit-types.md # detail loaded only when the agent needs it
The magic is progressive disclosure: the agent always sees the short description,
loads the SKILL.md body when the skill triggers, and reads bundled files like
references/commit-types.md only when the task calls for them — so a skill can carry a lot
of reference material at almost no cost until it's used.
---
name: conventional-commits
description: Write clear git commit messages… Use this whenever the user is committing code…
---
# instructions the agent follows once the skill loadsWant to build one? Start from templates/SKILL.md and follow
CONTRIBUTING.md.
FieldKit/
├── README.md # you are here
├── CONTRIBUTING.md # how to add a new skill
├── LICENSE # MIT
├── templates/
│ └── SKILL.md # annotated starter to copy
├── scripts/
│ └── package-skill.sh # zip a skill for claude.ai / API upload
└── skills/
└── conventional-commits/ # example skill
├── SKILL.md
└── references/
└── commit-types.md
A skill can direct an agent to run code and use tools, so install skills only from sources you trust, and skim a skill's files before you add it. Everything here is meant to be readable in a couple of minutes — that's the point. If you fork or extend these, keep them free of secrets and of anything that quietly reaches out to untrusted external sources.
MIT © 2026 Chris Stoklosa. Use these skills freely, and adapt them to your own workflow.