Fitcoach Pro banner
danilogj danilogj

Fitcoach Pro

AI community

Description

Prescription methodology for personal trainers as an AI skill for Claude and ChatGPT — with deterministic tools instead of model arithmetic.

Installation

This entry records only its repository, not the path inside it, so there is no exact command to give. Open the source below and copy the folder into ~/.claude/skills/, or the file into ~/.claude/agents/.

README

FitCoach Pro

**A prescription methodology for personal trainers, packaged as an AI skill — with the arithmetic taken away from the model.** Runs on Claude and on ChatGPT. Portuguese version: [`docs/README.pt-BR.md`](docs/README.pt-BR.md).

[![License](https://img.shields.io/badge/license-Apache%202.0-blue.svg)](LICENSE)

👋 Not a developer?

**[Read the trainer's guide](docs/TRAINER-GUIDE.md)** — installation with no terminal, a real conversation from first client to finished sheet, and a glossary. Written for people who have never installed anything like this. Em português: **[Guia do personal trainer](docs/GUIA-DO-PERSONAL.md)**.


Most "AI personal trainer" prompts fail the same way: ask for a program and you get twelve exercises, four sets each, no idea which gym the client trains at, no screening, and a body fat percentage quoted to one decimal place from a wrist sensor.

This repository is the opposite bet. It does not try to replace the trainer's judgment — it structures it. Screening before prescription. Split chosen from the client's real schedule, including the bad week. Exercises chosen from equipment that actually exists. Weekly volume audited per muscle group. Progression by an explicit criterion instead of vibes. And, throughout, an instruction the model does not follow on its own: **say when a number is an estimate.**

What makes it different

**The model never does the math.** Calories, macros, BMR, TDEE, weight trends, weekly set counts per muscle, 1RM, training load — every number comes from tested Python in [`tools/`](claude/fitcoach-pro/tools), not from the model's head. This is the difference between a prompt that warns about hallucination and a package that removes the opportunity for it.

**It refuses instead of guessing.** `metrics tdee-observed` exits with an error and "6 of the last 28 days have logged meals, this needs 10" rather than producing a confident number from nothing. Same for a rate of change under 14 days, or