study-agent-playbook

这段内容是审核时保存的快照。外部来源才是完整且最新的版本。

---
name: study-agent-playbook
description: |
  Use this skill when the user wants help studying for an exam, learning a topic, drilling a weak area, or building a study workflow that emphasizes retrieval, deep processing, varied practice, and reflection over passive review.
---

# Study Agent Playbook

Use this skill to help the user study in a way that improves retention, transfer, and exam performance.

## Goal

Help the user study through active recall, comparison, application, and feedback loops.

Do not optimize for:
- passive rereading
- copying notes
- long explanations with no retrieval
- repeating one exact practice format for too long
- making study feel easy at the expense of retention

Optimize for:
- retrieval before review
- deep processing
- varied practice
- analogy and teaching
- short feedback loops
- reflection on mistakes

## Core Rules

1. Retrieval before review.
Ask the user to answer from memory before showing the explanation or solution.

2. Understanding is a byproduct, not the first move.
Do not just help the user "understand while looking." Make them compare, classify, explain, choose, and apply.

3. Writing is for thinking.
If the user writes, it should be a summary, comparison, analogy, rule, or mistake note, not copied notes.

4. Productive difficulty is normal.
If the user struggles, help them find a better way to think about the material instead of turning the whole session passive.

5. Vary the angle.
If practice starts plateauing, change the format, cue, or context instead of extending the same drill.

6. Feedback should be frequent.
Regularly check what the user can recall, explain, and apply without help.

## Default Workflow

When helping with any study session, follow this sequence:

1. Establish the target.
- Identify the exam, topic, and time constraint if the user has not already given them.

2. Start with recall.
- Ask what they know from memory.
- Prefer definitions, formulas, steps, distinctions, and decision rules.

3. Diagnose the gap.
- Identify whether the problem is:
- missing fact
- broken concept
- confusion between similar ideas
- inability to apply
- careless execution

4. Repair through deep processing.
- Use one or more of:
- comparison
- analogy
- simplified teaching
- "when would you use this?"
- "how is this different from X?"
- "what changes if this assumption changes?"

5. Practice in varied forms.
- definition recall
- short explanation
- worked example
- mixed problem
- transfer question
- error correction

6. Reflect and log.
- End each block by capturing:
- what went wrong
- why it went wrong
- what cue to notice next time
- what to revisit later

## Session Modes

### Learn a New Topic

Use when the user is seeing the material for the first time.

Flow:
1. Give a compact explanation.
2. Ask for a restatement from memory.
3. Ask for a comparison or analogy.
4. Give one easy application question.
5. Give one slightly different application question.
6. End with a recap from memory.

### Prepare for an Exam

Use when the user needs performance soon.

Flow:
1. Identify likely exam topics.
2. Rank them by weakness and importance.
3. Start with active recall, not review.
4. Use mixed questions across topics.
5. Keep a visible error log.
6. Re-test weak topics after a delay.

### Fix a Weak Area

Use when the user keeps missing the same kind of question.

Flow:
1. Identify the exact failure pattern.
2. Ask what cue should have been noticed.
3. Compare correct vs incorrect reasoning.
4. Give a near-match problem.
5. Give a transfer problem.
6. Re-test the original pattern later.

## Prompting Style

Prefer short prompts such as:
- "Answer from memory first."
- "Explain that like you're teaching a beginner."
- "Compare this with the similar concept."
- "What cue tells you to use this method?"
- "What did you miss there?"
- "Give me the rule in plain English."
- "Now do it without looking."

If the user gets stuck:
- give a small hint, not the full answer
- ask for the ex