---
name: learner-diagnostic-skill
description: Diagnose the learner's goal, prior knowledge, misconceptions, constraints, and readiness before choosing instruction.
allowed-tools: [read_file, list_dir, grep_search, glob_files, research_note]
---

# Learner diagnostic

## Overview

Diagnosis establishes what the learner wants to do and what they can currently
demonstrate. It samples prerequisite knowledge and reasoning without turning the
first interaction into a full exam or assuming confidence equals competence.

## When to Use

Use at the start of a learning path, after a long gap, when a learner is stuck,
or when assessment evidence conflicts with the current objective map.

## Workflow

1. Clarify the learner's target capability, motivation, context of use,
   deadline, available material, accessibility needs, and preferred language.
2. Ask for a brief self-assessment, then treat it only as a hypothesis. Identify
   the smallest prerequisite set that would materially change the starting point.
3. Give one or two low-stakes diagnostic tasks that require recall, explanation,
   execution, or transfer rather than recognition alone.
4. Observe correctness, reasoning, strategy, confidence, help used, and error
   patterns separately. Distinguish a missing fact, faulty mental model,
   procedural gap, language issue, and careless slip.
5. Estimate mastered, uncertain, and untested prerequisites. Do not label a
   misconception from one ambiguous answer; probe it from a changed angle.
6. Produce a bounded diagnostic record and recommended starting objective,
   including uncertainty and what later evidence should trigger re-diagnosis.

## Verification

- [ ] The target capability and practical constraints are explicit.
- [ ] Starting level is based on observable work, not self-report alone.
- [ ] Error categories and uncertainty are distinguished.
- [ ] The recommended starting objective follows from prerequisite evidence.

## Red Flags

- Teaching before learning what the learner is trying to achieve.
- Using a long generic placement test unrelated to the target.
- Treating one wrong answer as a stable learner trait.