Write an app walkthrough script with AI
Plan the recording from the viewer’s side, with a line of narration beside each tap, pause or change of screen.
A good reason to write an app walkthrough script with AI is to discover a missing shot before everyone has recorded their voiceovers. The walkthrough-script-agent takes your app flow and lays it out on a timeline. If a confirmation only appears briefly, the person recording can see where to pause and which words should land over that screen.
You can work from a feature description, screenshots or authorised project files. The agent checks the feature inventory with you before drafting. A tutorial for one task can stand alone, or several short demonstrations can be joined into a master walkthrough.
What it does
- Map the steps a new customer takes to complete one task.
- Give a voice artist a continuous narration track to read.
- List the screens and actions the person recording needs to capture.
- Draft separate versions for onboarding and a shorter feature preview.
- Connect two demonstrations with a transition that explains where the viewer goes next.
Install and use
- Enable the skill and its writing roles in Claude Code. Screenshots and descriptions need no separate key. Existing recordings use the maker’s frame extraction helper with Python and FFmpeg.
- Ask, “Script a 40-second tour of finding an invoice. Here are the three screens in order. Put each line beside the action it describes.”
- Confirm the features, then read the narration aloud while following the steps in a test account. Adjust anything the current interface does differently.
FAQ
Source
Made by Michael Boeding. Original skill: walkthrough-script-agent, published on GitHub under the MIT licence.
View on GitHub