SLIDESHOT [docs] Markdown
For agents: read the markdown version of these docs →

Describe the demo flow

The single biggest lever on output quality is how you describe the workflow to your AI agent. Slideshot’s agent plans clicks, fills, waits, and navigation directly from that natural-language description, so a sharp prompt produces a sharp demo. A vague one produces a long, meandering recording.

What a good description looks like

A good description answers three questions:

  • Where does the demo start? Name the target URL and the screen the demo should land on.
  • What is the workflow? A step-by-step list of the actions the agent should perform.
  • When is the demo done? The success state. If you do not name an end state, the agent decides for you.

A repeatable structure

Paste this prompt into the AI agent that has the Slideshot skill or MCP server connected. The agent picks it up, fills in any blanks, and creates the run for you.

Record a demo video for [target_url] where you [first action], then
[second action], then [third action], and so on.

If you need to wait for something to load or animate, say "wait for two
seconds" between the relevant steps.

End the recording when [the success state, e.g. the report renders and
is fully visible].

Examples

A launch video:

Record a demo video for https://app.example.com where you create a new
project, invite a teammate by email, and open the analytics view.

Wait for two seconds when the analytics chart renders so the viewer can
see the result.

End the recording when the chart is fully visible and the success toast
has disappeared.

A docs walkthrough:

Record a demo video for https://app.example.com/settings/api-keys where
you create a new API key, copy it, and confirm the success state.

End the recording when the new key is listed and labeled.

Tips for sharper output

  • Use product language (“create a project”) instead of pure UI directions (“click the green button”). The agent follows the workflow even if the UI changes.
  • Name the end state. Slideshot’s editor trims aggressively after the last meaningful action, so a clear end produces a clean ending.
  • Mention private fields explicitly under “avoid showing”, or enable blur emails for runs against real data.
  • For workflows that need login, save credentials first (see Login to any web app) so the agent does not have to ask for them mid-run.
  • For demos that require an OTP or magic link, prepare for the run to pause in awaiting_input. See Runs.