Pick one:
# Isolated CLI install (recommended)
pipx install expert-mentor
# Plain pip
pip install expert-mentor
# From source, with the skill linked too
git clone https://github.com/voidstackloop/expert-mentor
cd expert-mentor
./install.shinstall.sh puts the mentor command on your PATH, links the skill into
~/.claude/skills/ and ~/.config/opencode/skills/, and runs a health check.
Undo it with ./uninstall.sh.
You can also run it without installing anything:
python3 scripts/expert_mentor.py --field "Rust"
./bin/mentor --field "Rust"Generate a reusable teaching system prompt:
mentor --field "organic chemistry" --level intermediateThe output is a complete system prompt. Paste it into any chat as the system prompt, or hand it to an API. To get a request body you can send directly:
mentor --field "contract law" --provider claude --emit json
mentor --field "Rust" --provider chatgpt --emit json
mentor --field "Rust" --provider ollama --emit modelfilementor run --field "Rust" --provider ollama --model qwen2.5:7bInside a session:
/resetclears the conversation history/exit(or Ctrl-C) quits- Token usage is printed after each turn on stderr
To use a cloud provider, set an API key first:
export ANTHROPIC_API_KEY=... # for --provider claude
export OPENAI_API_KEY=... # for --provider chatgpt
mentor run --field "Rust" --provider claudementor run --field "Rust" --provider claude --rememberThe learner is named from the field (rust). After the session:
mentor progress rust # see mastered / shaky / misconceptions
mentor review rust --apply # let the model update the profile from the transcript
mentor cards rust --generate # build flashcards
mentor quiz rust # review them with spaced repetition- Command reference for every option.
- Providers to configure Claude, ChatGPT, Gemini, or local models.
- How the mentor teaches to understand the pedagogy.