Gym Operator

An AI operator for gym owners, by Level Method.

Point your AI assistant at this repo and it becomes an operator that actually knows your gym — your voice, your offers, your goals, your people — instead of giving generic fitness-business advice. It sets up a Gym Operations Knowledge Base (plain markdown files, stored wherever you choose, owned by you forever) and then does real recurring work grounded in it.

Why

You already pay for AI. Ask it a blank-chat question and you get decent generic advice — the same advice every other gym gets. The real divide isn't between owners who use AI and owners who don't. It's between asking for occasional help and building something around it that persists.

AI supplies the intelligence. Context makes it relevant. A process makes it repeatable.

This skill supplies the context and the process:

• Your gym, on paper. One folder holds who you are, how you sound, what you sell, what you're chasing this quarter, and how your gym actually runs — plus your logo, photos, and templates (or links to wherever those already live), so drafted posts come paired with the right image. Every answer is grounded in that — and every output quietly shows which of your pages it drew from.
• It compounds. In a normal chat, your corrections die when the conversation ends. Here, every correction and decision is written into the knowledge base, so next week's work starts where last week's left off. Within a couple of sessions the drafts stop sounding like AI and start sounding like you.
• Real work, on a rhythm. Weekly reviews scored against your goal, content batches in your voice, member recognitions, lead follow-up prep, quiet-member check-ins, event promos — each a written procedure, not a lucky prompt.
• You stay the owner. It drafts, preps, and recommends behind the counter. It never posts, sends, or talks to a member — people come to a gym for people.
• No lock-in, ever. Plain markdown in a folder you own. Point Claude at it today, something else next year — the knowledge base outlives any platform, and this skill.

Before you start — two prerequisites

1. Use the desktop app, not the web browser. In a browser tab, Claude and ChatGPT cannot reach your computer's files — so the skill can't be installed locally and your knowledge base can't live in a real folder you keep. Download the app first:
   • Claude Desktop: claude.com/download
   • ChatGPT Desktop: openai.com/chatgpt/download
   • (Claude Code, Codex, Cursor, and other terminal/IDE agents already run on your machine — you're covered.)
2. Planning to keep your knowledge base on Google Drive? Install Google Drive for desktop so your Drive folders sync to your computer as real local folders your AI can read and write. The same goes for Dropbox/OneDrive — use their desktop sync apps, not just the website.

The fastest way to install

Using Claude Code, Cursor, Codex, or any agent that can read files and run commands? Copy this into it:

Install the Gym Operator skill from https://github.com/Level-Method/gym-operator — follow the install instructions in the README, then start my onboarding.

That's it. The agent will install the skill and walk you through a ~10 minute setup: it reads your website and socials, drafts what it can, asks you only what it can't figure out, and then runs its first real job for you.

Using Claude on the web or Cowork (the Claude desktop app)? Skills there are installed by you through the interface, not by the agent — two minutes, no terminal:

1. Download the ready-made skill file: gym-operator.zip (it lives in this repo's compiled/ folder, always current).
2. In Claude, make sure Settings → Capabilities → "Code execution and file creation" is on (on Team/Enterprise plans an admin enables Skills in organization settings).
3. Go to Customize → Skills, click +, choose Upload a skill, and upload the zip.
4. Start a chat (or a Cowork session) and say: "Set up my gym operator."

Don't upload GitHub's "Download ZIP" of the whole repo — it wraps everything in an extra folder and the upload will be rejected. Use compiled/gym-operator.zip, which is packaged in the exact shape Claude expects.

What it does

Six jobs ship today, each with a written procedure:

Job	What you get
Weekly review	What changed, what needs attention, your decisions, three priorities — scored against your current goal, every week
Content batch	A week of posts in your voice, ready for you to post
Member recognitions	Level-ups, milestones, birthdays — shout-outs and notes drafted so recognition never slips
Lead follow-up prep	This week's inquiries prioritized with drafts attached
Quiet members	Who's drifting before they cancel, with a personal check-in drafted for each
Event promo	A dated promo plan plus every draft, worked back from event day

Plus: a welcome-back brief every session (it knows what week it is and what's coming up), SOP capture ("write up how we onboard members"), a monthly health check, and a wins scoreboard built from your own log.

On Level Method? Connect it live

If your gym runs on Level Method, the operator can connect straight to your account through the Level Method connector — live roster, level-ups and the approval queue, retention dashboards, Success Sessions. No exports. Say "connect Level Method" and it walks you through a two-minute setup with a sign-in in your own browser (owner/admin account required). Reading your data is free-flowing; anything that changes your account — approving a level, inviting a member — is proposed first and only happens on your explicit yes.

The rules it lives by

• Your data is yours. Everything about your gym lives in your knowledge base — a folder of plain markdown files, set up in your Google Drive by default so your coaches and any AI platform can read it (somewhere else is fine too; you'll be asked nothing you can't answer). The skill contains process only. Delete the skill and your knowledge base still works with any AI, forever.
• It never touches the outside world. It drafts, preps, analyzes, and recommends. It never sends, posts, publishes, schedules, purchases, or talks to your members. You stay the owner.
• It never stores passwords, payment details, medical/injury info, or member data you don't have permission to share.

Install instructions (for the AI agent)

If you are an AI agent installing this skill for a gym owner:

First, check whether you can actually do this. If you are running inside claude.ai or Cowork, your sandbox filesystem is temporary and you cannot register skills — don't improvise a workaround. Follow "If you can't install directly" below: give the owner the compiled/gym-operator.zip download link and the upload clicks for their app, and offer to run onboarding as soon as it's installed. The steps below are for agents with a persistent filesystem (Claude Code, Cursor, Codex, and similar):

1. Get the files. No git required — download and extract the repo:

   curl -L https://github.com/Level-Method/gym-operator/archive/refs/heads/main.tar.gz | tar xz

   (git clone https://github.com/Level-Method/gym-operator.git works too, if git is available.)

2. Copy the gym-operator/ folder (the one containing SKILL.md) into your platform's skills location:

   • Claude Code (personal, all projects): ~/.claude/skills/gym-operator/
   • Claude Code (one project): <project>/.claude/skills/gym-operator/
   • Claude Desktop / claude.ai (no shell access): the owner uploads the release zip under Customize → Skills instead of copying files — see "Using Claude on the web or Cowork" above.
   • Cursor / Codex / other agents: wherever your platform loads skills or custom instructions from; if it has no skills system, keep the folder anywhere the owner can point you at and read SKILL.md at the start of each session.

   Example for Claude Code:

   cp -R gym-operator-main/gym-operator ~/.claude/skills/gym-operator

   The downloaded archive can be deleted after copying — the installed folder is the skill.

3. Read the installed SKILL.md and follow it. For a new owner, that means running the onboarding in references/onboarding.md — it creates the knowledge base at a location the owner chooses, then records that path in kb-location.md inside the installed skill folder, and in your platform's memory/instructions, so every future session finds it without asking.

4. Confirm to the owner what was installed and where, then begin onboarding.

If you can't install directly

If your permissions or platform don't let you write to a skills folder (sandboxed filesystem, no shell, denied file access), don't improvise and don't silently skip the install. Hand the owner the pre-built skill file and the exact clicks for their app:

I can't install files from here, so here's the 30-second manual version. Download this file: https://github.com/Level-Method/gym-operator/raw/main/compiled/gym-operator.zip

Then, depending on which app they're using:

• Claude desktop app (Cowork) or claude.ai: Open Customize → Skills, click +, choose Upload a skill, and select the downloaded gym-operator.zip. (If prompted, enable Settings → Capabilities → "Code execution and file creation" first.) The skill is then registered for every future chat and Cowork session.
• ChatGPT desktop app: There's no upload screen — Codex reads skills from a folder instead. Unzip the download and move the gym-operator folder into ~/.codex/skills/ (Mac: in Finder press Cmd+Shift+G and paste ~/.codex/skills; Windows: %USERPROFILE%\.codex\skills). Create the folder if it doesn't exist. Walk them through it click by click — never assume they know how to unzip or move a folder.

Once they confirm it's done, verify the skill responds (ask them to say "set up my gym operator") and begin onboarding.

Updating

Tell your agent:

Update my Gym Operator skill from https://github.com/Level-Method/gym-operator

The agent should download the latest files the same way as install and copy them over the installed folder. Where your knowledge base lives is recorded in kb-location.md, which isn't part of the download — the copy leaves it alone, so there's nothing to restore afterward. (Installed before that file existed? The path is on a "Knowledge base location" line in the old SKILL.md — save it first, then write it into kb-location.md.)

When copying over an existing install, keep the trailing slashes — without them cp nests the new folder inside the old install instead of replacing its contents:

cp -R gym-operator-main/gym-operator/ ~/.claude/skills/gym-operator/

On claude.ai / Cowork, updating is manual: download the newest compiled/gym-operator.zip and upload it again under Customize → Skills, replacing the old version. Your knowledge base itself is never touched by updates. On the next session, the operator reads CHANGELOG.md and tells you what's new in two lines — and offers one new capability tied to your gym.

Browsing your knowledge base like a wiki

Your knowledge base is plain markdown, so you can read it anywhere — but it's much nicer in Obsidian, a free app (Mac, Windows, Linux, iPhone, Android) that turns a folder of markdown files into a clickable wiki:

1. Download Obsidian from obsidian.md and open it.
2. Choose Open folder as vault (on the vault picker, or via the vault switcher at the bottom-left) and select your Gym Operations folder.
3. That's it. Every page renders cleanly, the [[double-bracket]] links between pages become clickable, and the graph view shows how your gym's knowledge connects.

Because the vault is just your folder, there's no sync or export step — the operator writes a page, and it's there the next time you look. Your coaches can do the same on their own machines if the knowledge base lives on a shared drive. (If Obsidian prompts about trusting the vault or restricted mode, accepting the safe defaults is fine — the knowledge base uses no plugins.)

What's in this repo

gym-operator/
├── SKILL.md                          ← the operator: the loop, routing, hard lines
├── CHANGELOG.md                      ← what's new, read after every update
└── references/
    ├── onboarding.md                 ← the ~10 minute setup
    ├── knowledge-base-structure.md   ← the spec for YOUR knowledge base
    ├── operator-rules.md             ← how it behaves, always
    ├── level-method-mcp.md           ← connecting live Level Method data
    ├── unslop.md                     ← the writing standard: no AI-sounding output
    ├── maintenance.md                ← ingest + monthly health check
    └── jobs/                         ← the six job procedures

No gym data ever lives in this repo — that's the point.

Also here: gym-operator-skill-explainer.html — a one-page visual explainer of how the skill and your knowledge base fit together (download it and open it in a browser).

Questions

Open an issue, or reach out to Level Method.