Lesson plan

Skill yugash007/edu-agent-skills/skills/onboarding/lesson-plan

Reusable educational skills for AI coding agents. Turn agents (Gemini CLI, Claude Code, Cursor, etc.) into Socratic mentors and active learning companions with a single command.

Install

npx -y skills add yugash007/edu-agent-skills --skill lesson-plan

Assembled from the repository path, not quoted from the project. Check it against their README if it does not work.

One thing to look at

4 stars4 stars. Stars are a popularity signal and not a quality one, but at this level it is likely that nobody has read this closely except its author, and you would be relying on your own review.

What its author says it does

Copied from the file, not written here

Use when structuring a multi-session learning roadmap with milestones, gates, and pacing calibrated to learner level and goal.

SKILL.md

3.5 KB, as published. Nobody here has run it

Purpose

Create a coherent learning roadmap that sequences sessions, defines milestones with verifiable gates, and prevents random topic jumping. Gives the learner a clear path from current level to stated goal.

Activation

Learner has a defined goal + time constraint ("interview in 2 weeks", "learn this repo in a month"). Multiple sessions planned with non-trivial sequencing. repo-understand produced a curriculum map needing personalization. Learner asks "where do I start and how do I get to X?"
Skip if: one-off session with no continuity. Learner needs immediate help on a specific concept/bug. Plan already exists and is current.
Routing: run repo-understand and find-your-level first if repo context or level is unknown. Revise plan when: learner stalls 2+ sessions, goal changes, or weak areas accumulate faster than planned.

Inputs

Learner goal (specific, time-bounded), level (from find-your-level), available sessions/time budget, known constraints (weak areas, learning style), repo/domain scope.

Plan Structure

Three layers: Phases (thematic blocks) → Sessions (individual units with focus + gate) → Gates (acceptance criteria before advancing).

Workflow

Decompose Goal — Restate as a specific, observable outcome (not "learn ML" but "train and evaluate a classifier on custom dataset"). Identify 3–5 sub-goals.
Check Prerequisites — For each sub-goal: list prerequisites, check which are confirmed, flag gaps as early sessions.
Define Phases — Group sub-goals into 2–4 phases. Each phase has a coherent outcome and takes 2–5 sessions. Order so each phase's output feeds the next.
Sequence Sessions — Break each phase into sessions. Each has: primary topic, skill(s) to use, gate criterion. Max 2 heavy concepts per session.
Set Gates — Measurable acceptance criteria per session: "can explain X mechanism," "passes challenge-generator at intermediate tier," "completes repo exercise." Gates prevent surface-level advancement.
Add Buffers — 1 revision session per 4–5 sessions. Define re-plan triggers: 2 consecutive gate failures → revise plan.
Get Buy-In — Present as a proposal. Ask: "Does this pacing feel right?" Adjust before finalizing.

Rules

DO: gates are non-negotiable — don't advance before gate is passed.
DO: prerequisite gaps always get sessions before dependent content.
DO: include buffer sessions and explicit re-plan triggers.
DO: present as proposal with learner input, not a mandate.
DON'T: pack more than 2 major concepts per session — cognitive overload kills retention.
DON'T: scope-creep beyond the stated goal — every session must link to a sub-goal.
DON'T: treat the plan as static — revise when real outcomes differ from projections.

Output

Responses should contain: goal + level + session budget, sub-goals, prerequisite status, phase breakdown (sessions with topics/skills/gates), buffer sessions, re-plan triggers, and proposal confirmation prompt. Format naturally.

Checklist

Goal is specific and observable.
Every session has an explicit gate.
No session has more than 2 major concepts.
Buffer sessions and re-plan triggers included.
Plan presented as proposal with learner confirmation.