Challenge generator

Skill yugash007/edu-agent-skills/skills/assessment/challenge-generator

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 challenge-generator

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 generating personalized practice challenges calibrated to the learner's weak areas, level, and project context.

SKILL.md

3.6 KB, as published. Nobody here has run it

Purpose

Generate targeted, level-appropriate practice challenges that force application over recall. Challenges must be grounded in learner's weak areas, current project, and chosen difficulty tier.

Activation

Concept just taught and practice needed. Learner asks for exercises/challenges. check-understanding or misconception-detector flagged a weak area. Interview/exam/milestone prep.
Skip if: no concept context established → run teach-concept first. Learner is blocked on production issue → debug-teacher.
Routing: pair with check-understanding to evaluate responses. Escalate to interview-mode for timed pressure practice.

Inputs

Target concept(s), learner level, known weak areas, project/repo context, preferred type (implement/debug/explain/design).

Challenge Types

Implement: write code from scratch to satisfy criteria.
Debug: identify and fix a deliberately broken snippet.
Explain: articulate behavior, tradeoffs, or mechanism in prose.
Design: propose architecture or algorithm for given constraints.

Difficulty Tiers

Beginner: one concept, well-defined, limited scope.
Intermediate: composite concepts, partially specified, tradeoff thinking required.
Advanced: ambiguous spec, production constraints, edge-case awareness required.

Workflow

Calibrate — Identify target concept(s) from session history or learner request. Select challenge type based on learning objective. Map learner level to difficulty tier.
Construct — State challenge clearly: context, constraints, success criteria, time/scope hint. Include starter scaffold where appropriate. Embed at least one non-obvious constraint testing deeper understanding.
Hint Ladder — Prepare 2–3 progressive hints (broad→specific) but don't volunteer them. Release only on request or after two failed attempts.
Evaluate — Grade reasoning quality, not just correctness. Identify what's right, where reasoning broke down, root cause. Classify: conceptual gap, implementation slip, or edge-case blindness.
Advance or Retry — Significant errors: simpler variant or targeted hint, then retry. Clean pass: increase tier or shift to next weak area.
Reinforce — Summarize the key insight the challenge surfaced. Record outcome for weak-area-tracker.

Rules

DO: require active reasoning — not definition recall.
DO: make success criteria explicit and testable before learner starts.
DO: state difficulty tier and confirm before starting.
DO: ground at least one challenge variant in the learner's current project/repo.
DON'T: reveal the solution before the learner attempts.
DON'T: release hints before at least one learner attempt.
DON'T: end without a one-sentence insight summary.
DON'T: generate generic challenges disconnected from session context.

Output

Responses should contain: context (concept + type + tier + grounding), challenge statement with acceptance criteria, starter scaffold if applicable, checkpoint prompt, and next step after evaluation. Format naturally.

Checklist

Difficulty tier stated and calibrated.
Acceptance criteria are measurable.
Challenge references session context or learner's project.
Post-challenge insight summary included.