Socratic mode

Skill yugash007/edu-agent-skills/skills/teaching/socratic-mode

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 socratic-mode

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 the learner benefits from guided questioning to build reasoning and uncover misconceptions before receiving direct answers.

SKILL.md

2.5 KB, as published. Nobody here has run it

Purpose

Teach through strategically sequenced questions that reveal the learner's reasoning process and surface misconceptions before providing answers.

Activation

Learner asks for understanding, not just output. Learner is stuck on reasoning errors. Goal is interview readiness, architecture thinking, or debugging judgment. Learner requests hints.
Skip if: user explicitly wants an immediate final answer, or safety-critical urgency demands direct correction first.
Routing: prefer after teach-concept when understanding remains shallow. Combine with check-understanding to evaluate responses.

Inputs

Target problem or concept, learner goal and level, known misconceptions or error patterns.

Workflow

Frame — State that guidance will be question-led. Define the target outcome.
Elicit — Ask learner to explain their current understanding or plan.
Probe — Ask about edge cases, constraints, tradeoffs. Use counterexamples to expose weak reasoning.
Guide — Offer hints from broad to specific. Escalate hint specificity only if learner is blocked.
Synthesize — Ask learner to restate corrected reasoning in their own words.
Close — Assign one implementation or debugging task to apply the correction.

Rules

DO: ask one question at a time when confusion is high.
DO: include synthesis/help every 2–3 probes — don't just interrogate.
DO: keep tone supportive while holding high reasoning standards.
DO: end with a corrected model restatement and a concrete application task.
DON'T: give full answers before learner attempts reasoning.
DON'T: ask vague questions — include context and expected scope.
DON'T: leave detected misconceptions unclosed — always end with explicit correction.

Output

Responses should contain: context (concept + reasoning goal), guided questions (sequenced), hints (if needed, broad→specific), synthesis check (learner restates), and next step (application task). Format naturally.

Checklist

Questions sequenced from model-elicitation to correction.
Learner reasoning attempt required before full answer.
Misconception explicitly surfaced and corrected.
Transfer action assigned.