Intelligent Tutoring Dialogue Designer Agent Skill

Script a multi-turn tutoring dialogue with branching responses for anticipated student difficulties.

Context cost
7k tokensestimated from the bundle, loaded when it triggers
Bundle
1 file · 27.8 kBtext throughout, nothing executable
Licence
not declarednone named by its author
Last change
—v1.0
Servers it uses
Noneruns standalone

What it does

Script a multi-turn tutoring dialogue with branching responses for anticipated student difficulties. Use when designing AI tutors, chatbot interactions, or structured one-to-one support scripts.

What this skill changes about the agent has not been set out here as a list. Its SKILL.md, shown below, says it in the author’s own words.

autotutorchidialoguegraessericapitstutoringvanlehn

The skill itself

A skill is instructions the model reads, so you can see what you are getting: below is the start of its SKILL.md. The front matter loads with every session, the body loads when the skill triggers, and the whole file comes with the install.

SKILL.md27.8 kB · preview
--- # AGENT SKILLS STANDARD FIELDS (v2) name: intelligent-tutoring-dialogue-designer disable-model-invocation: false user-invocable: true effort: medium # EXISTING FIELDS skill_id: "ai-learning-science/intelligent-tutoring-dialogue-designer" skill_name: "Intelligent Tutoring Dialogue Designer" domain: "ai-learning-science" version: "1.0" evidence_strength: "strong" evidence_sources: - "VanLehn (2011) — The relative effectiveness of human tutoring, intelligent tutoring systems, and other tutoring systems (meta-analysis)" - "Chi et al. (2001) — Learning from human tutoring (analysis of effective tutoring dialogues)" - "Graesser et al. (2005) — AutoTutor: An intelligent tutoring system with mixed-initiative dialogue" - "Chi & Wylie (2014) — The ICAP framework: linking cognitive engagement to active learning outcomes" - "Koedinger & Aleven (2007) — Exploring the assistance dilemma in experiments with cognitive tutors" input_schema: required: - field: "learning_objective" type: "string" - field: "anticipated_difficulties" type: "string" optional: - field: "student_level" type: "string" - field: "subject_area" type: "string" - field: "interaction_length" type: "string" - field: "student_model" type: "string" - field: "system_capabilities" type: "string" output_schema: type: "object" fields: - field: "dialogue_architecture" type: "object" - field: "dialogue_moves" type: "array" - field: "decision_rules" type: "object" - field: "example_dialogue" type: "object" chains_well_with: - "adaptive-hint-sequence-designer" - "self-explanation-prompt-designer" - "ai-feedback-design-principles" - "cognitive-tutoring-architecture-designer" teacher_time: "5 minutes" tags: ["tutoring", "dialogue", "ITS", "VanLehn", "AutoTutor", "Graesser", "Chi", "ICAP", "mixed-initiative"] ---
70# Intelligent Tutoring Dialogue Designer
71
72## What This Skill Does
73
74Designs the dialogue logic for an AI tutoring interaction — when to ask a question, when to give a hint, when to explain, when to prompt for self-explanation, and when to stay silent. This is the hardest design problem in intelligent tutoring: too much intervention prevents productive struggle and creates dependency; too little leaves students stuck and frustrated. VanLehn (2011) showed that the effectiveness of tutoring (human or AI) depends on the quality of the step-level interaction — systems that engage students in active reasoning at each step dramatically outperform systems that simply present content and evaluate final answers. Chi et al. (2001) analysed what makes human tutoring effective and found, counter-intuitively, that the most effective tutors were NOT the ones who explained the most — they were the ones who asked the right questions and created opportunities for students to self-explain. The output includes a complete dialogue architecture (the phases and branching logic of the interaction), a library of dialogue moves (the specific things the tutor can say or do), decision rules (when to use each move based on student responses), and a worked example showing the system in action. AI is specifically valuable here because it can sustain one-to-one dialogue at scale — but the dialogue must be deliberately designed or the AI will default to lecturing, which is the least effective tutoring strategy.
75
76## Evidence Foundation
77
78VanLehn (2011) conducted a comprehensive meta-analysis comparing human tutoring, intelligent tutoring systems, and no-tutoring conditions. He found that the critical factor was not WHO was tutoring but HOW. "Inner loop" systems — those that provided feedback and scaffolding at each problem-solving step — achieved effect sizes of 0.76, nearly matching human tutors (0.79). "Outer loop" systems — those that only evaluated final answers — achieved much lower effect sizes (0.31). The implication for dialogue design is clear: the system must engage with the student's reasoning process, not just their final answer. Chi et al. (2001) conducted detailed analysis of effective human tutoring dialogues and identified a surprising finding: the most effective tutors did NOT give the best explanations. Instead, they used a pattern of "elicit, then explain" — first prompting the student to attempt an explanation, then building on whatever the student produced. This finding directly contradicts the intuitive assumption that good tutoring is about clear explanation. The reason: when a tutor explains, the student passively receives. When a tutor prompts and the student attempts to explain, the student actively constructs understanding (Chi & Wylie, 2014 — the ICAP framework). Graesser et al. (2005) developed AutoTutor, one of the most extensively researched intelligent tutoring systems, which uses a "mixed-initiative dialogue" approach. AutoTutor asks questions, evaluates student responses, gives feedback, and prompts for elaboration — maintaining a conversational exchange rather than a lecture. Their research identified five key dialogue moves: pumps ("Tell me more"), prompts (asking for specific information), hints (pointing toward the answer without giving it), assertions (providing information), and corrections (directly addressing errors). Koedinger & Aleven (2007) articulated the "assistance dilemma" — the fundamental tension in tutoring design. Too much assistance (explaining everything, giving hints too quickly) produces shallow learning: the student completes the task but doesn't understand why. Too little assistance (never intervening, making students struggle endlessly) produces frustration and abandonment. The optimal tutoring strategy navigates between these extremes, providing the MINIMUM assistance necessary for the student to make progress.
79
80## Input Schema
81
82The teacher must provide:
83- **Learning objective:** What the student should master. *e.g. "Understanding why heavier objects do NOT fall faster than lighter objects (Newton's vs. Aristotelian physics)" / "Being able to identify the main argument in a non-fiction text and distinguish it from supporting evidence" / "Understanding that multiplying by a fraction less than 1 makes a number smaller, not bigger"*
84- **Anticipated difficulties:** Where students struggle. *e.g. "Students have the Aristotelian intuition that heavier = faster. They cite everyday experience (dropping a feather vs. a ball) as 'proof.' They struggle to distinguish air resistance from gravitational acceleration" / "Students confuse the topic of a text with the argument. They identify facts rather than claims. They struggle to distinguish what the author believes from what the author reports" / "Students apply the 'multiplication makes bigger' rule from whole numbers and are confused when ½ × 6 = 3"*
85
86Optional (injected by context engine if available):
87- **Student level:** Year group and proficiency
88- **Subject area:** The curriculum subject
This is a preview of SKILL.md. Download or install the skill to get the whole thing.
In the file
SKILL.md4,112 words
Files1
Licencenot declared
Why you can read it

Nothing in a skill executes. The client loads the text and the model follows it, so a skill can be audited the way a runbook is — by reading it.

What it costs in context

Skills are not billed by the call. They are paid for in context: every token the instructions occupy is a token your code, your diff and your conversation cannot use. Here is what this one takes and when it takes it.

≈820
always loaded
The name and description, so the model knows the skill exists and when to reach for it.
6,130
on trigger
The instruction body, read only when the skill fires.
3.5%
of a 200k window
Ten skills this size would take about 35% of the window before you open a file.
050k100k150k200k context window

7k tokens, estimated from the bundle at four bytes to the token, held for the rest of the session once it triggers. Heavy. Teams tend to install this one per project rather than globally, and load it only when the job comes up.

Servers bill, skills cost

A server charges by the month. A skill charges once per session, in context, and then keeps charging it for as long as the session lives.

Before and after

The same question, put to the same model twice: once as it comes, and once with these instructions loaded.

No worked example has been published for this skill yet.

The procedure it runs

The procedure has not been set out here step by step. It is in the body of its SKILL.md: the start of it is shown above, and the whole file comes with the install.

Prose, not code

These steps are written for a model to follow, not executed by a runtime. It can still be told to skip one, and it will say so when it does.

Servers it uses

None. This skill calls no MCP servers at all.

Everything it needs is in the instructions, so it works in a project with nothing connected — the model reads the file and changes how it works with what it can already reach.

Its SKILL.md does not declare which tools it asks the agent for, so what it touches is whatever the instructions themselves ask for — read them before you install it.

What it asks for
Writes filesnot declared
Network accessnot declared

This SKILL.md carries no allowed-tools line, so neither is declared. A skill grants no permissions of its own — it can only ask for tools your client already has.

What it will not do

Every skill is narrow, and the useful ones say where they stop. These are the jobs this one is the wrong tool for.

What this skill is not for has not been published here. Nothing is implied by that: it is a section the author has not filled in.

What is in the bundle

1 file, 27.8 kB on disk. A bundle is text throughout: the instructions the model reads, plus the templates it fills in.

  • SKILL.md27.8 kB
What is not in it

No dependencies and nothing executable: a skill is text the agent reads, so the bundle is 1 file you can review in full before installing.

Install

Installing copies the bundle into your project. Nothing runs at install time — the files sit on disk until the model reads them.

# Intelligent Tutoring Dialogue Designer · 7k tokens when loaded · macOS, Linux mkdir -p .claude/skills && t=$(mktemp -d) && curl -fsSL -o "$t/bundle.tar.gz" https://mcprush.com/api/skills/intelligent-tutoring-dialogue-designer/bundle.tar.gz && tar -xzf "$t/bundle.tar.gz" -C "$t" && [ -d "$t/intelligent-tutoring-dialogue-designer" ] && { [ ! -e .claude/skills/intelligent-tutoring-dialogue-designer ] || { mv .claude/skills/intelligent-tutoring-dialogue-designer "$t/.previous" && echo "Previous copy moved to $t/.previous"; }; } && mv "$t/intelligent-tutoring-dialogue-designer" .claude/skills/ && echo "Installed .claude/skills/intelligent-tutoring-dialogue-designer"
On Windows, or any OS with Node 20+
npx mcprush@latest skill add garethmanning/intelligent-tutoring-dialogue-designer

Windows PowerShell says 'running scripts is disabled on this system'? Run Set-ExecutionPolicy -Scope CurrentUser RemoteSigned once, or type npx.cmd instead of npx (codex.cmd, gemini.cmd, grok.cmd likewise). PowerShell 7 needs neither. Writes to .claude/skills/intelligent-tutoring-dialogue-designer/ in the current project. For every project instead, run mkdir -p ~/.claude/skills && t=$(mktemp -d) && curl -fsSL -o "$t/bundle.tar.gz" https://mcprush.com/api/skills/intelligent-tutoring-dialogue-designer/bundle.tar.gz && tar -xzf "$t/bundle.tar.gz" -C "$t" && [ -d "$t/intelligent-tutoring-dialogue-designer" ] && { [ ! -e ~/.claude/skills/intelligent-tutoring-dialogue-designer ] || { mv ~/.claude/skills/intelligent-tutoring-dialogue-designer "$t/.previous" && echo "Previous copy moved to $t/.previous"; }; } && mv "$t/intelligent-tutoring-dialogue-designer" ~/.claude/skills/ && echo "Installed ~/.claude/skills/intelligent-tutoring-dialogue-designer" (macOS, Linux) — or add --global to the npx line. The curl line replaces .claude/skills/intelligent-tutoring-dialogue-designer/ only after the download succeeds, so files left there by another skill of the same name, or by an older release, do not mix in. A copy already there is moved aside, not deleted, and the line prints where.

Which clients pick it up on their own

A skill is a folder of text. Most of these read a skills folder and pick it up without being told; the rest take the same folder another way — an upload in the app, a parameter on an API call — and then choose it by its description just the same.

Claude Code.claude/skills/
Claude DesktopCustomize › Skills
ChatGPTSkills › Create › Upload
Cursor.cursor/skills/
VS Code.github/skills/
Codex CLI.agents/skills/
Gemini CLI.gemini/skills/
Grok Build.grok/skills/
Microsoft CopilotSkills › Upload a skill
PerplexityComputer › Skills › Upload a skill
DeepSeek Harness.dsh/skills/
Zed.agents/skills/
Devin Desktop.devin/skills/
Windsurf (legacy).windsurf/skills/
OpenAI Agents SDKInstructions or sandbox Skills
HTTP / APIa request parameter
This release
Version1.0
Publishedno release date on file
PriceFree
Referencegarethmanning/intelligent-tutoring-dialogue-designer

Versions

v1.0 is what is on the shelf; no release here carries a date. Instructions change more often than APIs do — a skill can be rewritten entirely without anything it depends on moving.

v1.0
  • No earlier releases have been published to the marketplace.
Pinning

The install command always fetches the release that is current the day you run it — v1.0 today — and cannot ask for an earlier one. The copy on your disk does not change until you run it again.

Reviews

—no reviews yet

Nobody has reviewed this skill yet. The rating is the mean of the reviews written here, so there is none until somebody writes the first.

Sign in to review it

A review is tied to an account so it can be answered, edited and taken down by the person who wrote it.

Sign in
Who can post

Anybody with an account, one review each. It carries the version that was current when you wrote it, and the publisher may reply once.

Publisher
Servers0
Claim this skill