Classify Lesson Types

Skill: Classify Lesson Types

Description

Walks through each SLT in a set and interviews the user to determine the appropriate lesson type. Builds a classification document and refines heuristics about when each lesson type applies.

Lesson Types

| # | Type | Inputs | Best For | |---|------|--------|----------| | 1 | Product Demo | SLT + screenshots | UI walkthroughs, platform features, "click here then here" sequences | | 2 | Developer Documentation | SLT + code + docs link | API usage, library integration, code-heavy technical skills | | 3 | How To Guide | SLT + optional materials | Step-by-step procedures, workflows, processes without heavy UI or code | | 4 | Organization Onboarding | SLT + org context | Org-specific setup, policies, team conventions | | 5 | Exploration | SLT + framing questions | Big ideas, thought leadership, worldview challenges, conceptual territory |

Instructions

Path Resolution

Resolve file paths based on your execution context:

• Plugin context (${CLAUDE_PLUGIN_ROOT} is set): Read knowledge from ${CLAUDE_PLUGIN_DATA}/knowledge/ (user data), falling back to ${CLAUDE_PLUGIN_ROOT}/knowledge/ (seed data). Read research from ${CLAUDE_PLUGIN_ROOT}/knowledge/research/.
• Clone/symlink context (default): Read knowledge from knowledge/ relative to the project root (research is at knowledge/research/).

Pre-Execution Knowledge Check

Before classifying SLTs, read the knowledge base for heuristics that should improve your initial guesses. If any knowledge file does not exist, skip it and proceed without prior heuristics — rely on the lesson type definitions below.

1. Read knowledge/lesson-types/heuristics.yaml

• Use verb_patterns to inform initial classification
• Use subject_patterns to refine based on topic
• Note confidence levels — high confidence patterns can be stated more firmly

1. Read knowledge/lesson-types/edge-cases.yaml

• Check if this SLT matches a known ambiguous pattern
• If so, skip to the discriminating_question that resolved it before

1. Surface heuristics to user (if relevant patterns exist):

```markdown ### Classification Insight

Based on previous courses:

• "explain" verbs → Exploration (87% of the time)
• API/library subjects → Developer Documentation (92% of the time)
• This SLT matches a known edge case — asking the same clarifying question that resolved it before.

```

---

The user will provide a markdown file containing SLTs. Read the file, then for each SLT:

1. Present the SLT and Your Initial Read

For each SLT, share:

• The SLT text
• Your initial guess at lesson type (with brief reasoning)
• One or two clarifying questions that would help confirm or revise

2. Interview with Discriminating Questions

Ask questions that help distinguish between candidate lesson types. Good discriminating questions:

Product Demo vs. How To Guide:

• "Does this SLT require showing the actual UI, or could it be taught with written steps alone?"
• "Is the learner clicking through an interface, or following a conceptual procedure?"

Developer Documentation vs. How To Guide:

• "Is code the primary artifact the learner produces, or is code incidental to a broader workflow?"
• "Would a code-free learner still benefit from this SLT?"

Exploration vs. How To Guide:

• "Is there a 'right answer' or correct procedure, or is this about developing perspective?"
• "Does this SLT ask the learner to adopt a stance or challenge assumptions?"

Organization Onboarding vs. others:

• "Does this SLT only make sense within a specific organization's context?"
• "Would a generic version of this SLT exist, or is it inherently org-specific?"

3. Capture the Decision and Reasoning

After the user responds, record:

• The chosen lesson type
• Key factors that determined the choice
• Any edge cases or "it could also be..." notes

4. Build Heuristics

As you work through the SLT set, notice patterns:

• Which SLT verbs tend toward which lesson types?
• What subject matter clusters with each type?
• Where do ambiguities arise, and how were they resolved?

Output Format

After classifying all SLTs, produce a markdown document:

# Lesson Type Classification: [Course Name]

## Summary

| Lesson Type | Count | SLTs |
|-------------|-------|------|
| Product Demo | [n] | 1.2, 2.1, ... |
| Developer Documentation | [n] | ... |
| How To Guide | [n] | ... |
| Organization Onboarding | [n] | ... |
| Exploration | [n] | ... |

## Classifications

### Module 1: [Module Name]

#### SLT 1.1: "[SLT text]"
- **Lesson Type**: [type]
- **Key Factors**: [why this type fits]
- **Edge Notes**: [any ambiguity or "could also be..." notes]
- **Inputs Needed**: [what context/assets this lesson will require]

[Repeat for each SLT]

---

[Repeat for each module]

## Heuristics Developed

Patterns observed during classification that can guide future SLT sets:

### Verb Patterns
- SLTs with "[verb]" tend toward [lesson type] because...

### Subject Matter Patterns
- [Topic area] SLTs tend toward [lesson type] because...

### Ambiguity Patterns
- When an SLT could be [type A] or [type B], we resolved by asking...

## Context Shopping List

Assets needed to build lessons, organized by type:

### Screenshots Needed (Product Demo)
- [ ] [description of screenshot for SLT X.Y]

### Code Examples Needed (Developer Documentation)
- [ ] [description of code for SLT X.Y]

### Framing Questions Needed (Exploration)
- [ ] [topic area for SLT X.Y]

### Org Context Needed (Organization Onboarding)
- [ ] [context needed for SLT X.Y]

Guidelines

• Interview, don't dictate. Your initial guess is a starting point for conversation, not a verdict.
• One SLT at a time. Don't batch — the interview loses value if rushed.
• Capture reasoning, not just labels. The heuristics section is as valuable as the classifications.
• Flag hybrid cases. Some SLTs genuinely span types. Note these rather than forcing a single label.
• Build the shopping list as you go. Each classification implies inputs; capture them immediately.
• Refine heuristics across sessions. If this skill is run multiple times, the heuristics section should grow more precise.

Source & license

This open-source skill is cataloged on AgentStack and links to its original source — we do not rehost the code.

• Author: Andamio-Platform
• Source: Andamio-Platform/coach
• License: MIT

Install and usage instructions live in the source repository linked above.