Session 0: Experience — Witness the Power of AI Agents Firsthand

Session Overview

Preparation (For Instructors)

Required Environment

1. GitHub Repository (prepare a template repository)

   • Example repository name: ai-agent-workshop-demo
   • All participants pre-invited as Collaborators
   • GitHub Actions configured to run Claude Code

2. Files to Include in the Repository

   ai-agent-workshop-demo/
   ├── CLAUDE.md              # Instructions for the AI agent
   ├── .github/
   │   └── workflows/
   │       └── claude-agent.yml  # GitHub Actions configuration
   ├── output/                # Directory for deliverables
   └── README.md              # Repository description

3. Example CLAUDE.md Contents

   # Project Configuration
   
   You are a business research assistant.
   Execute tasks assigned via GitHub Issues and save deliverables to the output/ folder.
   
   ## Working Rules
   - Create deliverables in Markdown format
   - Use the file naming convention `YYYY-MM-DD-task-summary.md`
   - Always cite sources for research
   - Create a Pull Request upon completion and reference the Issue

4. GitHub Actions Configuration File (claude-agent.yml)

   • Triggers Claude Code execution based on Issue labels or comments
   • Refer to Appendix A for detailed configuration

Rehearsal Checklist

• Create a test Issue and confirm that Claude Code runs successfully
• Confirm that a PR is automatically created and deliverables are saved to output/
• Confirm that all participants’ GitHub accounts have access to the repository
• Confirm that the Wi-Fi connection is stable
• Confirm that the demo projector/screen sharing works

Timetable

1. Opening (10 minutes)

Suggested instructor script:

“Today, we’ll start by experiencing firsthand how the way we work is about to change. There are no complicated steps involved. All you need is the web browser you use every day. I have just one request — when what you see today surprises you, let yourself feel that surprise. Don’t hold back.”

What to do:

• Brief self-introduction
• Ask with a show of hands: “How many of you have used an AI chatbot before?”
• Set expectations: “What you’ll experience today is not a chatbot”

2. Demonstration (15 minutes)

The instructor performs a live demo.

Step 1: Create an Issue

Share your screen showing the GitHub repository and create an Issue with the following content:

Title: Research on the Latest Trends in Remote Work Productivity

Body:
Research the latest survey data and trends on remote work and hybrid work
from 2024–2026, and create a report covering the following perspectives:

1. Key statistics (impact on productivity, employee satisfaction, and turnover rates)
2. Case studies of successful companies (approximately 3)
3. Future outlook

Target audience: Managers in the corporate strategy department

Step 2: Show the AI agent in action

• Use the GitHub Actions logs and Issue comments to show that the AI agent is working
• Explain: “Right now, the AI agent is gathering information from the web and writing a report”
• While waiting for the task to complete, briefly explain the big picture with the next slide

Step 3: Review the deliverable

• Share your screen when the PR is created
• Review the quality of the deliverable (Markdown report) together
• Emphasize: “This was created in just a few minutes”

Suggested instructor script:

“What you just saw was not AI simply answering a question. It received a task, conducted its own research, created a file, and even submitted it for review. This is what an ‘AI agent’ is.”

3. Hands-On Experience (40 minutes)

Participants create their own Issues and assign tasks to the AI agent.

Step 1: Choose a Task (5 minutes)

Have participants choose from the following task templates (free-form input is also welcome):

Template A: Industry Research “Identify the three latest trends in the XX industry and summarize each with an overview and business impact in approximately 500 words.”

Template B: Competitive Analysis “Compare the services of Company XX and Company YY, and create a comparison table covering features, pricing, and target customers.”

Template C: Proposal Draft “Create a draft proposal for an internal study session on XX. Include the purpose, target audience, proposed agenda, and required preparations.”

Step 2: Create the Issue (10 minutes)

• Each participant navigates to the GitHub repository
• Click “New Issue”
• Create an Issue based on the chosen task template, tailored to their own work
• The instructor supports any participants who get stuck

Instructor Tips:

• Encourage concise Issue titles with details in the body
• Explain that including specifics like “who the document is for” and “how long it should be” leads to better results

Step 3: Review the Deliverable (15 minutes)

• Once the AI agent completes the task, guide participants to review the PR
• Have them read the deliverable and evaluate its quality
• Ask them to leave their impressions or revision requests in the PR comment section

Step 4: Experience Requesting Revisions (10 minutes)

• Demonstrate that the AI agent responds to revision requests written in PR comments
• Examples: “Please add more specific data,” “Please reorganize this as a table”
• Let participants experience the cycle of “request → execution → review → revision request”

4. Reflection and Discussion (20 minutes)

Discussion prompts:

1. What surprised you: “What was the most surprising thing about this experience?”
2. Possibilities: “Can you think of situations in your own work where this could be useful?”
3. Concerns: “Did you feel any concerns or anxieties after trying this out?”

Key points for the instructor to summarize:

• AI Agent ≠ AI Chatbot

  • Chatbot: A single exchange of question → answer
  • Agent: Task → planning → execution → deliverable submission (autonomously performs multiple steps)

• GitHub is the “AI agent’s office”

  • Issue = Task assignment
  • PR = Deliverable submission and review
  • Comments = Feedback and revision requests
  • In other words, it has the same structure as assigning work to a human team member

• Why you should learn this now

  • AI agents are not just a “tool” — they represent a new way of working
  • You need the ability to give proper instructions and evaluate deliverables
  • This skill is not just for engineers

5. Preview and Closing (5 minutes)

Suggested instructor script:

“Today, you experienced everything just by clicking buttons on GitHub. Starting next session, we’ll move into understanding how AI agents work and learning to use them with full control. In the next session, we’ll explore ‘What is an AI agent?’ and try running Claude Code on your own PC.”

Homework (optional):

• Write down three tasks in your daily work that you think could be delegated to an AI agent
• Browse the GitHub repository and check out other participants’ Issues and PRs

Appendix A: Example GitHub Actions Configuration File

# .github/workflows/claude-agent.yml
name: Claude Agent

on:
  issues:
    types: [opened, labeled]
  issue_comment:
    types: [created]

jobs:
  agent:
    if: |
      (github.event_name == 'issues' && contains(github.event.issue.labels.*.name, 'agent-task')) ||
      (github.event_name == 'issue_comment' && contains(github.event.comment.body, '@claude'))
    runs-on: ubuntu-latest
    permissions:
      contents: write
      pull-requests: write
      issues: write
    steps:
      - uses: actions/checkout@v4

      - name: Run Claude Code
        uses: anthropics/claude-code-action@v1
        with:
          anthropic_api_key: ${{ secrets.ANTHROPIC_API_KEY }}
          github_token: ${{ secrets.GITHUB_TOKEN }}

Appendix B: Troubleshooting

Appendix C: Instructor Checklist (Day-of)

• GitHub Actions is enabled on the repository
• ANTHROPIC_API_KEY is set in the repository’s Secrets
• Verified operation with a test Issue
• Final confirmation of participant list and GitHub invitation status
• Projector/screen sharing verified
• Wi-Fi connection stability confirmed
• Backup mobile hotspot prepared (optional)