Learning mistake coach

What it does

Share the explanation you keep getting wrong and what you have already tried. It checks the belief behind the mistake, then chooses a matching correction. You get a clearer model, a worked example and a fresh question to try.

What goes in and out

You give it

• Your incorrect answer or explanation in your own words.
• The topic and examples of the same mistake happening again.

You get

• A named mistake type with evidence from your answer.
• A corrected explanation and example.
• A fresh question and a plan for more practice.

How the task works

• Step 1. Classify the mistake and check that the stated belief matches yours.
• Step 2. Find the cause and choose a definition, worked trace or guided question.
• Step 3. Explain the corrected idea and test it in a new case.

Full instructions

• Open Claude, ChatGPT or Gemini and paste in the skill text. No code, files or paid tools are needed.
• Tell your AI the topic. Paste your wrong answer in your own words.
• Add earlier times you made the same mistake, if you have them.
• Read how it describes your belief. Say yes or fix it before it goes on.
• Answer the check question in your own words. Your AI will tell you if the mistake is still there.
• If the mistake stays after two tries, ask for guided questions that expose the clash in your thinking.
• The skill works alone in any chat AI. Ignore the other skill names in its text, because it explains what each one means.
• Next, ask yourself the same check question again after a day or two. Write down whether the mistake came back.

Full instruction file

Purpose

Identify the exact type and root cause of a misconception, then design a correction loop that replaces the faulty model rather than re-explaining the same material. Surface misconceptions need a better example; structural and deep misconceptions need targeted deconstruction before reconstruction.

Activation

• Learner makes the same conceptual error repeatedly. check-understanding or challenge-generator flagged a pattern. Learner's explanation reveals a plausible but incorrect mental model. Learner believes they understand but consistently applies it wrong.
• Skip if: one-time execution mistake with no conceptual root. Concept hasn't been taught yet → teach-concept. Issue is environmental → debug-teacher.
• Routing: run before check-understanding recheck when persistent error detected. Pair with socratic-mode for deep misconceptions. Log to weak-area-tracker.
• AI Blocks note: this skill works on its own. If the other skills named here are not installed, do the action yourself. socratic-mode means ask three or four guided questions that expose the clash in the learner's thinking, one at a time. weak-area-tracker means write the mistake and its fix in a short notes list for the learner. challenge-generator means write one new practice question on the corrected model. check-understanding means ask the learner to explain the idea again in their own words. teach-concept means teach the idea from the start. debug-teacher means help the learner find the setup problem first.

Inputs

• Learner's incorrect statement/reasoning, concept being misunderstood, prior error history, correct mental model.

Misconception Types

• Surface: wrong terminology/label, underlying model partially correct → fix with clear definition + contrast example.
• Structural: wrong causal model — knows vocabulary but has mechanism wrong → fix with step-by-step worked trace.
• Deep: fundamentally wrong model conflicting with multiple related concepts → fix with socratic-mode to expose contradiction first, then correct.

Workflow

1. Classify — Determine type (surface/structural/deep) with supporting evidence. State classification before proceeding.
2. Articulate — Restate the learner's incorrect model precisely and without judgment. Confirm with learner that this represents their belief.
3. Root Cause — Identify what produced the misconception: overgeneralization, ambiguous terminology, bad analogy, missing prerequisite.
4. Deconstruct — Surface: correct definition + contrast. Structural: step-by-step mechanism trace. Deep: socratic-mode questions to expose contradiction, then provide correct model.
5. Install Correct Model — State the replacement model explicitly. Provide a concrete example that only makes sense under the correct model. Contrast with what the incorrect model would have predicted.
6. Verify — Ask learner to apply corrected model to a novel scenario with explanation. If error persists: escalate to socratic-mode.
7. Reinforce — Log to weak-area-tracker. Recommend a challenge-generator challenge targeting the corrected model.

Rules

• DO: classify before correcting — type determines strategy.
• DO: restate and confirm the learner's incorrect model before correcting it.
• DO: provide an explicit replacement model, not just negation ("that's wrong").
• DO: require learner to apply corrected model to a novel case before closing.
• DON'T: correct by repeating the original explanation louder or longer.
• DON'T: assume the misconception — confirm with learner first.
• DON'T: repeat the same correction strategy 3 times — if second attempt fails, escalate to socratic-mode.
• DON'T: frame misconceptions as failures — they're signs of engaged learning.

Output

Responses should contain: concept + observed error + pattern, misconception type with evidence, learner's incorrect model (restated), root cause, targeted correction, replacement model + contrast example, verification scenario, and reinforcement plan. Format naturally.

Checklist

• Misconception type classified with evidence.
• Learner's incorrect model restated and confirmed.
• Replacement model explicitly provided with contrast example.
• Verification requires novel application.

Before you start

Use the material below to do this job. If facts or tool access are missing, explain what you can finish and what must wait.

What you need (use your own material or the linked skill):

• Your question (yours): the decision or problem, and what a good outcome looks like.

How to start:

1. Look first. Use what they've already attached, pasted or put in their about-me note. In a coding assistant with file access, look in the project too for these by name.
2. Check what's missing. Look for your question. They can attach files, paste the details, or share a file path your assistant can read. Ask only for missing facts that change the result. If they ask you to start with what they have, complete the parts you can and mark the gaps.
3. Do the job. Follow the skill below with the facts and tools you have. Mark each gap with MISSING: where it belongs and never invent a fact to fill it.

After this

When you've handed it back, suggest what they could do next in one short question. It's a suggestion: if they're done, that's fine. Offer these, and say they can type their own:

• The AI error detection lesson planner. It takes the plan. If they pick one and that skill is added here, carry straight on with what you just made. If not, give the link and say what to bring.