As if that wasn’t enough, the next day, 24-Feb-2026, Bart publicly unveiled his demo repo:
A collection of zsh scripts that demonstrate every major feature of swiftDialog through an interactive, self-guided tour.
Quick Start
# clone this repo
git clone https://github.com/swiftDialog/demo.git swiftDialog-demo
# or download the zip
curl -sL "https://github.com/swiftDialog/demo/archive/refs/heads/main.zip" -o swiftDialog-demo.zip && unzip swiftDialog-demo.zip -d swiftDialog-demo
cd swiftDialog-demo
./run_demos.zsh
Later that same day, the demo repo’s first Pull Request was merged and closed.
Inspiration + AI
Beginning about the middle of March 2026, I was away from my home office for a dozen consecutive days both receiving and conducting training.
While in this environmental state-of-flux — finding coding more challenging than normal — I received some heavenly inspiration:
Train AI using the demo repo
“Brilliant!” I thought. While I couldn’t easily code, AI didn’t care about the comfort level of the hotel bed.
Playground
The playground directory offers guided, step-by-step walkthroughs for using AI to create swiftDialog scripts. It includes five realistic Mac Admin scenarios — such as onboarding forms, compliance prompts, app deployment, policy acknowledgments, and Inspect Mode workflows — along with exact prompts, expected AI responses, troubleshooting tips, and quick-reference cards for common patterns.
Think of it as your hands-on training ground: follow the exercises to learn how to effectively “teach” an AI assistant the style and conventions of the official demo repo.
AI Skills
The skills directory contains ready-to-use instruction packs that turn large language models into swiftDialog experts. Each skill provides curated guidance, references, templates, and repository-specific conventions so the AI reuses patterns from the demo suite instead of guessing.
Currently available:
claude-swiftdialog-builder — Optimized for Claude Sonnet (uses CLAUDE.md)
Both packs produce clean, repo-aligned scripts that follow best practices like mktemp-based temporary command files and content-aware window sizing.
One-shot Prompt
Review skills/claude-swiftdialog-builder/CLAUDE.md and use it to create an internal-use swiftDialog shell script for our team. The script should display on-screen instructions and require the user to explicitly acknowledge them before proceeding. Treat this as an internal workflow, not a public-facing demo.
#!/bin/zsh
# Internal: Policy Acknowledgment Dialog
# Displays on-screen instructions and requires explicit acknowledgment before proceeding.
DIALOG="/usr/local/bin/dialog"
# --- Guard: swiftDialog must be installed ---
if ! command -v "$DIALOG" &>/dev/null; then
echo "Error: swiftDialog not found at $DIALOG" >&2
exit 1
fi
# --- Acknowledgment Dialog ---
# The 'enableButton1' modifier on --checkbox unlocks button1 when the box is checked.
# --button1disabled sets it locked on open; both flags are required together.
"$DIALOG" \
--title "Before You Proceed" \
--message "Please read the following instructions carefully before continuing.\n\n**Step 1:** Ensure you are connected to the corporate VPN.\n\n**Step 2:** Close any open applications that may be affected by this workflow.\n\n**Step 3:** Do not interrupt this process once it has started.\n\nIf you have questions, contact the Help Desk before proceeding." \
--icon "SF=exclamationmark.shield,colour=#FF9500" \
--checkbox "I have read and understood the instructions above,name=acknowledged,enableButton1" \
--button1text "Acknowledge and Continue" \
--button1disabled \
--button2text "Cancel" \
--moveable \
--width 700 \
--json 2>/dev/null || exit 0
# --- Workflow continues here after acknowledgment ---
echo "User acknowledged. Proceeding with workflow..."
Support
Community-supplied, best-effort support is available on the Mac Admins Slack (free, registration required) #swiftdialog channel, or you can open an issue.
Your Mac already has AI. Apple ships a language model with macOS -apfel unlocks it with one brew install. No downloads, no API keys, no config. The fastest path to local AI.
Dan K. Snelson
