Process Guide

AI Teaching Process Guide

This page is not centered on a single final `skill.md` file. It teaches how teachers can understand, sequence, and practice skill usage through workflow thinking: define the task, choose the right skill type, build checkpoints, interpret the meaning of each skill entry, and revise the process until it becomes teachable.

Workflow-first Learning process Teaching design Reusable practice

Focus

What this guide teaches

The goal is not to memorize tool names. The goal is to understand how teaching tasks should be analyzed and matched to the right kind of skill support.

Principle

Process before product

A polished slide deck or handout can hide weak reasoning. This guide emphasizes the path behind the output: task framing, source preparation, tool pairing, validation, and revision.

Build A Skill

How to build a skill that is teachable and reusable

Start from repeated work, not from format

A skill usually starts because the same kind of task keeps returning: making slides, organizing a handout, cleaning data, finding references, or checking whether an AI explanation is accurate. That repetition is the signal that the task can be compressed into a reusable method.

In practice, a skill is a packaged teaching method. Its value is not that the text looks sophisticated. Its value is that someone else can reuse the logic with less confusion.

Write these four things first

  1. What task the skill is meant to handle.
  2. What inputs the user will typically provide.
  3. What a good output should look like.
  4. What signs show that the result is still weak and needs revision.

If these four points are unclear, the skill will likely remain a private note rather than a reusable teaching tool.

1

Define the task

State whether the skill is mainly for documents, slides, data work, literature support, or validation.

2

Define the input

List the materials users usually bring, such as topic, data, PDF, references, audience, or time limit.

3

Define the output

Describe what a useful result should look like: structure, pacing, format, or evidence quality.

4

Add validation

State how the result will be checked for readability, teachability, reproducibility, or source accuracy.

The first version does not need to be complete, but it must be testable

A practical first draft should be small. Run it against one real task, note where the instructions are vague, which inputs tend to break the flow, and what outputs still feel weak. Then feed those lessons back into the skill.

Mature skills carry visible traces of revision

A strong skill usually shows why certain constraints were added, which warnings came from failure, and which checkpoints were introduced to stop repeated mistakes. That is why teaching a skill should include how it was shaped, not only what the final file says.

Learning Stages

Four stages for learning skill-based teaching workflows

1

Identify the task

Decide whether the teaching need is explanation, presentation, data interpretation, or live demonstration.

2

Match the skill type

Choose document, slide, spreadsheet, literature, validation, or orchestration support based on the task instead of starting with a fixed tool.

3

Build checkpoints

Separate each workflow into input, processing, output, and review so the method can be taught clearly.

4

Revise intentionally

Learning improves when each run is reviewed for clarity, efficiency, and error detection, not just for completion.

Generic Skill Types

How to understand each skill in a transferable way

Skill typeGeneric roleWhat the learner should practiceCommon mistake
`presentations`Turns content into live teaching rhythm and visual pacingPractice selection, sequencing, and slide-level signal controlTreating slides as a full handout
`documents`Turns concepts and procedures into readable teaching materialPractice structure, explanation depth, and conversion from spoken process to written guidanceFocusing only on formatting instead of instructional clarity
`spreadsheets`Organizes fields, calculations, comparisons, and chart logicPractice data cleaning, column design, interpretation, and translation into teaching languageJumping to presentation before analysis is ready
`pdf`Extracts reusable knowledge from existing materialsPractice summarizing, regrouping, and rewriting for new audiencesUsing the original PDF as the final teaching product
`crossref-apa-bibliography`Builds the research and citation backbonePractice source retrieval, comparison, and citation consistencyListing references without showing how they support the teaching argument
`openai-docs`Provides official capability and product groundingPractice verification, version awareness, and source checkingTreating informal online claims as official documentation
`external-agent-orchestrator`Plans tool roles and sequence across a workflowPractice decomposition, coordination, and return-path validationLooking only at the final result instead of tool timing and role assignment
`execution-recovery-guardrails`Turns failure points into teachable validation rulesPractice checkpoint design and error recovery logicSeeing failure as a side issue instead of part of the teaching process

Skill.md Guide

Readable interpretation of skill.md content

Instead of pasting raw configuration text, this section translates the core instructional meaning behind each skill entry: what kind of task it handles, what usually goes in, what comes out, and what a teacher should learn to notice while using it.

presentations:Presentations

Used for turning a teaching topic into a live-facing slide structure.

Typical input

Topic, audience, time limit, examples, or data points.

Typical output

A slide deck with clear pacing, section order, and message-per-slide discipline.

Teaching checkpoint

The learner should ask whether each slide carries one primary teaching signal.

documents:documents

Used for turning steps, concepts, and examples into reusable written teaching material.

Typical input

Instructional steps, case explanations, workshop notes, or procedural guidance.

Typical output

A structured handout, guide, plan, or explanatory document.

Teaching checkpoint

The learner should check whether the text still works when the teacher is not present to explain it.

spreadsheets:Spreadsheets

Used for score tables, survey scales, schedules, and educational data handling.

Typical input

Raw columns, test scores, attendance, rubric data, or comparative records.

Typical output

Clean fields, summary tables, charts, and interpretable comparisons.

Teaching checkpoint

The learner should verify that the data is already interpretable before moving it into a talk or handout.

pdf:pdf

Used for extracting reusable ideas from existing PDF-based material.

Typical input

Research papers, slide exports, handouts, or report appendices in PDF form.

Typical output

Summaries, extracted sections, restructured content, or teaching-ready rewrites.

Teaching checkpoint

The learner should check whether the PDF has been transformed into teachable material instead of merely copied.

crossref-apa-bibliography

Used for research support and citation management in scholarly sharing.

Typical input

Topic keywords, DOI, author names, or journal references.

Typical output

APA references, source lists, and a clearer citation backbone.

Teaching checkpoint

The learner should know which references actually support the teaching claim being made.

openai-docs

Used for official capability checks, API guidance, and product-grounded explanations.

Typical input

Model questions, product comparisons, API usage, or official feature lookups.

Typical output

Traceable official explanations and lower-risk teaching statements.

Teaching checkpoint

The learner should practice verifying claims before presenting them as fact.

external-agent-orchestrator

Used for multi-tool sequencing, role assignment, and validation flow design.

Typical input

Complex tasks, multiple tool roles, verification rules, and cost constraints.

Typical output

A clearer workflow showing who does what, in what order, and where results are checked.

Teaching checkpoint

The learner should be able to explain the role of each tool, not just show the end product.

execution-recovery-guardrails

Used for turning repeated mistakes into teachable recovery logic.

Typical input

Failure cases, repeated errors, broken steps, or unstable process points.

Typical output

Checklists, recovery rules, and a more stable repeatable workflow.

Teaching checkpoint

The learner should ask whether failure has been converted into a rule others can follow.

Scenario Paths

How workflows shift across teaching scenarios

  • Teacher workshops start with activity design, then split roles between slides and handouts.
  • Academic talks start with evidence and literature, then compress into presentation form.
  • Assessment analysis starts with field cleaning and comparison structure before interpretation.
  • AI demonstrations start with capability verification, then move into tool order and checkpoints.

Teacher Questions

Questions worth repeating during practice

  • Am I teaching a concept, a step, or a full workflow?
  • Where would a participant most likely get stuck when doing this alone?
  • Which part needs a chart, and which part needs explanation?
  • Am I showing a product, or am I showing the reasoning behind the product?

Practice Cycle

Suggested learning loop

First run

Build the smallest workable version. One page, three slides, or one analysis view is enough.

Second run

Add validation. Check sources, sequence, chart meaning, and whether the workflow can be followed by others.

Third run

Rebuild the same topic for a different setting so the learner practices transfer instead of repetition.

Common Errors

Three errors that weaken learning

  1. Treating the skill as the content instead of the method for shaping content.
  2. Showing polished outputs without exposing the judgment and revision steps.
  3. Learning one template for one scenario and failing to transfer it elsewhere.

Teaching Aim

What the learner should take away

The intended outcome is a way of thinking: identify the task, choose the right support type, sequence the workflow, set checkpoints, and refine the result through repeated practice.

YT

Author

Author information

This page was organized and authored by Yu-Ting Shih. Its focus is AI teaching communication, instructional workflow design, research-oriented educational sharing, and the transferable use of skill-based methods across teacher workshops and academic presentations.