lesson-scaffold

name: lesson-scaffold description: Restructures external learning materials into a conceptual scaffold tailored to the learner's current level. Reads learning state (current-state, goals, session logs) to classify concepts and shape the scaffold. Writes a persistent scaffold file to learning/scaffolds/. Use when the user provides learning materials and wants to understand the "why" before executing.

Lesson Scaffold

Restructure procedural material by conceptual dependency, shaped to the learner's current state. The scaffold is a map, not a lecture.

Inputs

1. Source material (required). URL, file path, or pasted text — any learning content.
2. Context (optional). Energy, time, emotional context, what they need. Shapes scope.

Phase 1: Gather

Read local learning state. Fail fast and degrade gracefully — every source is optional.

Step 1: Learning state check

Check whether learning/current-state.md and learning/goals.md exist.

• Both exist: Full gather. Proceed normally.
• Some exist: Work with what's available. Silent.
• None exist: Tell the user: "No learning profile found — I'll scaffold from the material alone. Run /intake for a personalized scaffold." Proceed with concepts classified as new.

Step 2: Recent session context

List files in learning/session-logs/. For the most recent 1-2 log files, read the YAML frontmatter. Extract:

date: YYYY-MM-DD
project: project-name
concepts:
  - name: concept-name
    score: N
    gap: type
arcs:
  - arc-name

Also scan the body of the most recent log for "remaining work" or "next steps" — material that connects to unfinished threads gets extra relevance.

From the collected frontmatter, build a picture of:

• Recent concepts and scores — what was just reviewed and where gaps remain
• Arcs touched recently — which developmental lines are active
• Unfinished threads — remaining work that the material might connect to

If no session logs exist, skip silently.

Step 3: Learning state files

If the learning state files exist, read them:

learning/current-state.md — concept inventory. Extract:

• Concept scores — focus on concepts relevant to the source material
• Gap types — conceptual vs. procedural vs. recall informs how to scaffold each concept
• Evidence sources — quiz-verified scores carry more weight than estimates (see .claude/references/scoring-rubric.md)

learning/goals.md — active goals. Extract:

• Goal aspirations and capabilities required
• Which goals the source material serves

Reference files consulted as needed:

• .claude/references/scoring-rubric.md for score interpretation
• .claude/references/developmental-model.md for ordering heuristic

Phase 2: Analyze and Scaffold

Step 1: Fetch the source material

Read or fetch the source material provided by the user.

Step 2: Extract concepts

Extract every concept the material teaches or assumes — explicit and implicit. Implicit concepts (things the lesson takes for granted) are often where the real learning edges are.

Step 3: Classify each concept using gathered state

For each extracted concept, assign a classification:

• Solid (score ≥ 4 in current-state) — use as anchor. Bridge new concepts to these. Brief mention only.
• Growing (score 2-3) — reinforcement opportunity. Note the gap type: conceptual gaps need framing, procedural gaps need reps, recall gaps need spaced exposure.
• New (not in current-state) — new territory. Connect to the nearest known concept as a bridge.
• Prerequisite gap (score ≤ 1, or missing but the material assumes it) — flag as risk. Suggest pre-work or flag for extra attention during execution.

If no learning state exists, classify all concepts as new.

Step 4: Connect to goals

If learning/goals.md exists, note:

• Which goals does this material serve?
• Which arcs does it touch?
• One sentence connecting the material to the learner's trajectory.

Step 5: Produce the scaffold

Six sections, each shaped by the concept classifications:

What this material covers — one paragraph connecting the material to the learner's goals and current state. What it teaches, why it matters now.

Concepts to understand — each concept with:

• Classification (solid / growing / new / prerequisite gap)
• A one-sentence bridge to something the learner already knows
• For growing concepts: note the gap type
• For prerequisite gaps: flag the risk

Execution sequence — the instructor's steps reordered by conceptual dependency. Group steps that share a concept. Mark which concepts each step exercises. Weight time toward growing and new concepts — solid concepts need less attention.

Resurfaced gaps — low-scoring concepts from current-state.md that this material happens to touch. Spaced repetition for free.

Goal connection — one line: what this material moves forward in the learner's trajectory.

Pre-work (only if prerequisite gaps were found) — specific concepts to review or brush up on before starting.

Step 6: Write the scaffold file

Write the scaffold to learning/scaffolds/YYYY-MM-DD-<topic-slug>.md.

Derive the topic slug from the source material title, URL path, or filename. If the learning/scaffolds/ directory doesn't exist, create it.

File format:

---
date: YYYY-MM-DD
source: <URL, file path, or "pasted text">
arcs: [list of arcs this material touches]
concepts: [list of concepts extracted]
---

# Scaffold: <topic>

## What this material covers
<paragraph connecting to learner's goals>

## Concepts
<classified concept list with bridges, gaps, and status>

## Execution sequence
<reordered steps grouped by concept, weighted toward growth edges>

## Resurfaced gaps
<low-scoring concepts this material touches>

## Goal connection
<one line: what this moves forward>

## Pre-work
<only if prerequisite gaps found>

Step 7: Link to session log

Check whether learning/session-logs/YYYY-MM-DD.md exists for today.

If it exists: add a scaffold: field to its YAML frontmatter pointing to the scaffold file.

If it doesn't exist: create it with minimal frontmatter:

---
date: YYYY-MM-DD
scaffold: scaffolds/YYYY-MM-DD-<topic-slug>.md
---

This creates a breadcrumb so session-review can connect the scaffold's predictions to session evidence. Session-review will merge its own frontmatter (project, concepts, arcs) into this file later — don't duplicate those fields here.

Step 8: Present and confirm

Present the scaffold to the user. Ask whether the scope and classification feel right. If they adjust, update the file.

They drive.

Graceful Degradation

Boundaries

• Don't execute the assignment. Scaffold for understanding, not walkthrough.
• Don't update learning state. This skill writes to learning/scaffolds/ and session log frontmatter only. Session-review owns all state updates.