Skills

Ready-made, agent-neutral recipes for building on the NxVET API

Last updated: July 23, 2026

Overview

A skill is a ready-made recipe for something useful you can build on top of NxVET. You don't write any code — you give the recipe to an AI assistant on your computer, and it builds a working little app for you, asking simple questions as it goes.

The recipes work with whichever assistant you prefer — Claude Code, Codex, Cursor, or another — and nothing is hidden: you (or your assistant) can read exactly what each one does before running it. New to all this? Jump to the Quick Start.

Quick Start — zero to a working app

You don't need to be a programmer. The idea: you get an AI assistant that works on your computer, give it a NxVET key, and then paste one message that tells it which skill to build. It does the rest — downloading the skill, setting things up, and asking you questions in plain language as it goes. About 15 minutes.

Step 1 — Get an AI assistant on your computer

These skills are built by an AI coding assistant such as Claude Code (from Anthropic, the makers of Claude) or Codex (from OpenAI). Follow their install page — it walks you through it — and sign in with your account. If you get stuck here, email us and we'll help you get set up.

Step 2 — Get your NxVET key

A "key" is like a password that lets the assistant read your clinic's NxVET data. Log in at app.nx.vet, open Integrations → API Keys, and create one. Copy it somewhere safe and keep it private — you can switch it off any time from that same page. (More detail: Authentication.)

Step 3 — Pick a skill and paste the message

Make a new empty folder for your app, open your assistant there, and paste the message from the skill you want. Every skill in the gallery has a “Copy the starter message” button that gives you the exact wording. For the voice-notes skill it looks like this:

Please build the NxVET "voice-notes-to-reminders" skill for me.
Download it from https://api.nx.vet/skills/voice-notes-to-reminders.zip,
read the SKILL.md inside, and set it up step by step. I'm not technical,
so explain things simply and ask me for my NxVET key when you need it.

That's the whole trick — the assistant reads the instructions itself and does the technical parts for you.

Using Claude Code? You can instead install the skill once as a plugin and then call it by name — handy if you'll use it again. In Claude Code, run:

/plugin marketplace add nervextech/nxvet-skills
/plugin install voice-notes-to-reminders@nxvet

then start it anytime with /voice-notes-to-reminders:voice-notes-to-reminders. Either way, skip to Step 4.

Step 4 — Answer a few questions

As it builds, the assistant will ask you things like your NxVET key, which device to watch, and your time zone. Your key and everything the app produces stay on your own computer — NxVET only ever sees the voice note, never your calendar or reminders. When it's done you'll have:

  • a daily reminders note you can open and read, and
  • calendar files you double-click to drop an event into Outlook or Google Calendar.
Good to know: the same steps work for every skill in the gallery, and you can run more than one (make a separate folder for each). Prefer to do it by hand? The step-by-step recipe is in each skill's SKILL.md — see Install for your agent.

Install for your agent

Skills are plain files — load them into whichever agent you use. Full per-agent notes ship in each bundle's INSTALL.md; here's the short version.

# Option A — install as a plugin, then invoke it by name (recommended).
# In Claude Code, run:
/plugin marketplace add nervextech/nxvet-skills
/plugin install voice-notes-to-reminders@nxvet
#   then invoke it anytime with:
/voice-notes-to-reminders:voice-notes-to-reminders
#   ('/plugin marketplace update nxvet' pulls new skills + updates later)

# Option B — no install: work in a project folder and say:
#   "Read SKILL.md and implement it. Start with Phase 0."
#   (this is what the Quick Start "starter message" does for you)

# Optional: add the NxVET MCP server so the agent calls the API as tools
claude mcp add nxvet --transport http https://mcp.nx.vet/mcp \
  --header "Authorization: Bearer nxvet_sk_YOUR_API_KEY"
# Codex reads AGENTS.md from the working directory:
cd voice-notes-to-reminders
printf '%s\n' '# Build task' \
  'Read SKILL.md in this folder and implement it. Start with Phase 0.' > AGENTS.md
codex

# Optional MCP server — ~/.codex/config.toml:
# [mcp_servers.nxvet]
# command = "npx"
# args = ["-y", "nxvet-mcp"]
# env = { "NXVET_API_KEY" = "nxvet_sk_YOUR_API_KEY" }
# Load the recipe as a Cursor rule, or just open SKILL.md and prompt:
mkdir -p .cursor/rules
cp voice-notes-to-reminders/SKILL.md .cursor/rules/voice-notes.md
#   then: "Read SKILL.md and the reference/ files, then implement it,
#          phase by phase, starting with Phase 0."

# MCP-capable editors: add https://mcp.nx.vet/mcp with a
# "Authorization: Bearer nxvet_sk_..." header (see /mcp.html).
# No skill machinery required. Paste SKILL.md (and the reference/*.md files)
# into the chat and say:
#
#   "Build this as described, phase by phase, starting with Phase 0.
#    Ask me for the API key when you reach it."
#
# The two helper scripts are stdlib-only Python 3 — the model can reproduce
# or adapt them from the bundle.
Key safety. Whichever agent you use, the API key belongs in a local .env file only — never hard-coded in a script, never committed to a repo, never pasted into a chat that leaves your machine. Revoke it instantly from app.nx.vet/integrations if it ever leaks.

For AI agents

A machine-readable index of all skills lives at https://api.nx.vet/skills.txt (llms.txt-style). It lists each skill with a description and direct URLs to its SKILL.md, bundle, scripts, and reference notes, so a connected assistant can discover and fetch a recipe on its own. It's also linked from llms.txt and llms-full.txt.

If your agent is connected to the NxVET MCP server, point it at skills.txt and it can pull a skill and start building against the live API tools.