Arcade Source Content

Arcade Source Content

Overview

Turns raw product material + named subtopics + a demo list into one scene file per subtopic, all sharing identical variable names so Arcade can rebuild the same visual template with different copy.

Core principle: the product layer is written once and repeated verbatim; only the subtopic layer varies. Everything is driven by the demos.

Inputs

Ask for whatever is missing. Do not proceed on assumptions.

Process

1. Write the product layer — once

Derive from raw content, then reuse byte-identical in every file: Scene 1 Title, Scene 1 Subtitle, Scene 2 Kicker, Word 1/2/3, Scene 2 subtitle, closing TITLE, closing Subtitle.

Only Top left corner changes between files.

Before writing any file, draft these seven values as one block and check them against every subtopic — they must work for all of them, not for the subtopic you happen to be writing. If a value only fits one subtopic, it belongs in the subtopic layer.

2. Assign demos to subtopics

A demo may serve more than one subtopic — reuse is fine and expected. A demo that fits nowhere goes in the coverage report as unused; do not force it into a subtopic.

3. Count key points from demos

Key points = demos assigned to that subtopic, capped at 4, minimum 1. One demo, one key point. When a subtopic has 5+ demos, pick the strongest 4 and list the rest as surplus in the coverage report.

Scene numbering follows the count:

A 2-demo subtopic emits 2 pillars and closes on Scene 6. Renumbering the closing scene is correct; a 7-scene file for a 2-demo subtopic is not.

4. Pick the Scene 2 demo

Scene 2's demo previews the subtopic. Use the subtopic's broadest demo, and prefer one that is not already a key point demo. When every demo is already a key point demo (common at 1–2 key points), reuse the strongest and note it in the coverage report.

When one demo is a clear overview of the subtopic rather than a specific capability ("Automation for AIOps" next to three named integrations), reserve it for Scene 2. It becomes the preview and does not count toward the key point total from step 3.

A demo with no link is not usable. Exclude it from key point counting and list it in the coverage report under unusable, so the count reflects what can actually ship.

5. Write the subtopic layer

Per key point, from the raw content — not from the demo description alone: a Kicker, a Title, and exactly 3 bullets.

6. Emit files

<product-slug>-source-content/
  00-coverage.md
  01-<subtopic-slug>.md
  02-<subtopic-slug>.md

Current directory unless the user gives a path.

Field Contract

Every field, every file. template.md in this directory is the structure to copy; example.md is the quality bar to match. Read both before writing.

Match the capitalization style the user used for the product name.

Bullets escalate

Three bullets, three jobs, in order:

1. Capability — what it does
2. Mechanism — how it works
3. Outcome — why it matters to the buyer

Bullet 1: Let AI identify the best fix; AAP carries it out consistently.
Bullet 2: Integrate via event-driven triggers or direct API calls.
Bullet 3: Position AAP as the trusted execution layer beneath any AI stack.

Every bullet traces to a claim in the raw content or a demo description. Do not invent metrics, customer names, or capabilities.

Coverage report

00-coverage.md is part of the deliverable, because handoff decisions get lost in chat:

• Table: subtopic → key point count → demos used
• Demos used in more than one subtopic, and where
• Unused demos, with a one-line note on why none fit
• Anything inferred rather than sourced, marked <!-- REVIEW --> in the file it appears in

Common Mistakes

Before handing off

Check each generated file:

• Product-layer fields are identical across all files (diff them)
• Scene count = key points + 4, closing scene numbered correctly
• Every Pillar N desc matches its scene's Title verbatim
• Every DEMO is name — link and the link was supplied by the user
• Every key point scene has exactly 3 bullets
• Top left corner present on every scene
• No placeholder or TODO text

Next step

These files are the input to arcade-booth-generate, which turns one of them at a time into a generated Arcade video.