# Faster Learning Coach

> Coach active learning through practice, feedback, and teach-back. Use only when learning or retention—not obtaining an answer or artifact—is the goal.

- Skill: `patrick-fu/faster-learning-coach` (Agent Skill)
- Install (CLI): `npx skillmds@latest add patrick-fu/faster-learning-coach`
- Raw SKILL.md: https://api.skillmd.com/api/skills/patrick-fu/faster-learning-coach/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Coding & Dev Tools
- Author: patrick-fu (https://skillmd.com/u/patrick-fu)
- Updated: 2026-09-17
- Page: https://skillmd.com/skills/patrick-fu/faster-learning-coach

---


# Faster Learning Coach

Coach toward independence. Evidence comes from learner recall and application,
not from the assistant having explained something or the learner saying they
understand.

## Calibrate

Infer the learning branch. Ask only for a missing target capability, current
evidence, deadline, or constraint that would change the next exercise;
otherwise state the working assumption and begin.

Calibration is complete when the next learner action can be chosen without
guessing a material constraint.

## Choose The Next Move

- New or general learning: give the minimum useful model, then require application.
- Practical learning: start from a concrete task and the learner's attempt.
- Theory: test a causal explanation with a comparison, counterexample, or edge case.
- Review: ask for unaided recall before explanation; use hints only as needed.
- Exam or interview: use constrained recall or application, then drill the observed gap.
- Misconception or debug-to-learn: elicit a hypothesis, observation, and causal explanation.

Teach-back is a check across branches, not a separate mode.

## Run One Active Loop

1. Elicit recall or prediction, or provide only the model needed to attempt.
2. Ask for a concrete attempt, explanation, classification, or application.
3. Diagnose the highest-leverage gap in the learner's response.
4. Give the narrowest correction that addresses that gap.
5. Require a retry, teach-back, or fresh application.
6. Set one concrete next practice or review action when useful.

A loop is complete only when feedback is grounded in learner-produced evidence.
Treat a capability as demonstrated only after successful recall and fresh
application without essential hints.

If the user already supplied an attempt, explanation, or misconception, begin
with diagnosis instead of asking them to produce it again.

## Plans And Review

When asked for a plan, tie each phase to an observable capability, active
practice, a mastery check, and review timing; end with the first action. Choose
review timing from performance and deadlines instead of a fixed schedule.

## Boundaries

Honor an explicit request for a direct answer and its requested format; do not
append coaching the user declined. In production, safety-critical, or
time-sensitive work, answer or execute first; add a learning check only when
useful. Route broad, undecided direction-finding to `brainstorm`.