majiayu000/spellbook / skills/build-product-demo
build-product-demo
VerifiedPlan, produce, or diagnose evidence-backed product demo videos and screen-recorded promotional walkthroughs. Use when the user asks to make a product demo, launch video, feature showcase, app walkthrough, demo reel, or polished recording; wants every important capability shown without a slow feature tour; asks to remove dead time or fix narration-to-action pacing; or needs a repeatable script, capture plan, and verified final media. Do not use for fictional commercials with no real product proof, general-purpose video editing, or documentation-only tutorials that do not need a promotional narrative.
| File | Kind | Size |
|---|---|---|
| SKILL.md | skill md | 13.0 KB |
| agents/openai.yaml | config | 239 B |
| evals/evals.json | config | 5.0 KB |
| references/artifact-contract.md | markdown | 6.8 KB |
| references/directing.md | markdown | 6.6 KB |
| references/production.md | markdown | 6.0 KB |
| scripts/analyze_demo_pacing.pyexecutable | script | 15.5 KB |
| scripts/probe_demo_media.pyexecutable | script | 6.8 KB |
| scripts/validate_demo_plan.pyexecutable | script | 16.3 KB |
references/directing.md
View on GitHub# Directing Product Proof
Use this reference after product inspection and before writing a timed script.
## 1. Choose the decision spine
Write one sentence that can reject attractive but irrelevant footage:
`This demo turns [audience doubt] into [confidence] by showing [observable product transformation].`
Cascade it through the production:
| Layer | Decision test |
|---|---|
| Feature selection | Does this capability materially answer the doubt? |
| Action | Does the product visibly change state or produce evidence? |
| Framing | Can the audience locate the cause and result without searching? |
| Narration | Does it add value, consequence, or context the screen cannot? |
| Sound | Does it redirect attention or register a result? |
| Cut | Does the next image add, compare, compress, or resolve information? |
## 2. Calibrate against a reference
Choose one concrete artifact before designing. Compare it on:
- native product surface versus explanatory graphics;
- first meaningful product action;
- observable events per ten seconds;
- longest unchanged frame and longest audio void;
- shot-scale, crop, and color progression;
- real operator behavior, friction, response, and consequence;
- sound continuity and landing energy.
Extract two or three qualities to preserve. A genre label such as “Apple-like,”
“cinematic,” or “high energy” is not calibration.
## 3. Design the change
Define:
```text
Opening product state:
Opening audience belief:
Obstacle or doubt:
Turning proof:
Landing product state:
Landing audience belief:
Next action or call to action:
```
If the opening and landing states are equivalent, the demo is structurally
flat. A collection of unrelated finished screenshots does not create change.
## 4. Select proof, not coverage
Use three depths:
1. **Core proof**: show the complete cause, product action, and consequence.
2. **Differentiator**: show why the product is meaningfully different.
3. **Supporting evidence**: establish trust quickly without breaking the flow.
An exhaustive feature list may still be delivered as a coverage appendix. It
should not dictate equal screen time.
For each feature ask:
- What doubt does it remove?
- What exact product state changes?
- What can the camera or terminal actually capture?
- What evidence proves it is not a staged animation?
- What is the shortest truthful action chain?
## 5. Build chapters and beats
Use chapters when the user goal, proof question, or visual condition changes.
Use beats when at least one changes:
- the operator adopts a new tactic;
- the product accepts a meaningful input;
- the audience receives decisive information;
- the product produces or persists a result;
- an error, gate, or recovery path redirects attention;
- the immediate question changes.
Record the audience state before and after every beat. Remove a beat when those
states are identical and the product state does not change.
## 6. Hook with native action
For a promotional demo, prefer a legible result in the opening seconds, then
return to the credible starting state. The hook should create a question the
body can answer, not spend the best result before the audience understands it.
Useful hook shapes:
- before/after transformation;
- failure or friction followed by the finished result;
- one-command input followed by a surprising real output;
- rapid result montage from one coherent user journey;
- a high-stakes question immediately answered by product evidence.
Avoid company history, logo animation, setup narration, or permission dialogs
as the first meaningful image unless one of them is the product's value.
For a promotional software demo, require a meaningful native product action by
five seconds. A claim card about a result does not satisfy this requirement.
## 7. Control attention
Give every shot one primary job. Use the smallest device that makes it legible:
- stable wide frame for orientation;
- crop or zoom for a decisive detail;
- cursor path only when cause depends on pointer location;
- highlight or callout for a result that would otherwise be missed;
- chapter card when the proof question changes;
- split screen only for a genuine comparison;
- contact sheet or result montage only after the relationship is established.
Do not stack zoom, cursor halo, caption, callout, and motion background around a
simple result. Visual hierarchy should become clearer as proof accumulates.
Count observable events inside the shot: input, response, reveal, result,
motivated crop, sound cue, or cut. During the hook, aim for a new event every
0.8–2 seconds; during the body, every 1.5–3 seconds. Do not count ambient drift,
blinking cursors, progress spinners, or decorative particles.
## 8. Align narration and action
Assign each channel a job:
- **screen**: how the product behaves;
- **narration**: why the action matters or what changed;
- **on-screen text**: names, metrics, commands, or conclusions that must remain
inspectable;
- **UI/action sound**: cause, completion, or state transition;
- **music**: larger pressure and release, never missing product logic.
Start the corresponding action before or shortly after narration creates the
expectation. Land on the result as the sentence resolves. If processing takes
time, show a readable state change, compress it transparently, or redesign the
proof path.
## 9. Shape pace
Choose `max_information_gap_seconds` for the audience and format rather than
copying a universal number. Useful starting points:
| Format | Starting range |
|---|---:|
| 15–60 second promo | 3–6 seconds |
| 60–180 second product demo | 5–8 seconds |
| Longer technical walkthrough | 8–15 seconds |
This limit governs meaningful information, not arbitrary cuts. A moving cursor
or animated spinner is not new information. A held result is valid when the
audience needs time to read, compare, judge, or absorb it; record that reason in
the beat plan.
## 10. Reject template signals
Redesign when several of these appear together:
- equal-duration beats unrelated to action length;
- repeated centered headlines and symmetrical cards;
- abstract benefit copy where behavior should be visible;
- a single dark-gradient palette with decorative neon accents;
- no cursor, typing, state transition, failure, latency, or consequence;
- product output retyped into a fake terminal or presentation layer;
- identical shot scale and composition across changing claims.
## 11. Review the sequence
For each planned cut complete:
`Cut from [state/action/sound] to [new state/action/sound] because the audience now needs to [learn/feel/locate/compare/anticipate].`
Reject cuts that only decorate, and reject long takes whose internal action
does not develop the proof.
Related skills
| Skill | Stars | Updated | Works with | Status | Install |
|---|---|---|---|---|---|
| rea-changelog-updatemorluto/reaPrepare or rewrite REA release changelogs and GitHub release notes from pinned Git history, with verified contributor thanks and Release Please synchronization. | Stars: 472k | Updated | Claude CodeCodex | Verified | Install |
| academy-guideanthropics/skillsStop and check this skill before finishing any reply to a question about how to use Claude or a Claude product — it recommends matching courses, tutorials, and use cases from Claude Academy (academy.claude.com), Anthropic's learning hub. Trigger on: "how do I", "how can I", "getting started with", "what can Claude do", "teach me", "learn to use"; questions about artifacts, projects, skills, plugins, connectors, MCP; requests about rolling Claude out to a team, class, or organization; and any ask for training materials, onboarding content, or learning resources. Use it when the user is learning how to use a feature or product — not when they are mid-task and just want the task done. This skill composes with other skills: after consulting product documentation to answer how a Claude feature works, also check here for a matching course or tutorial — a docs-grounded answer and an Academy recommendation belong together. Only recommend on a strong match; never invent Academy content. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |
| claude-apianthropics/skillsReference for the Claude API / Anthropic SDK — model ids, pricing, params, streaming, tool use, MCP, agents, caching, token counting, model migration. TRIGGER — read BEFORE opening the target file; don't skip because it "looks like a one-liner" — whenever: the prompt names Claude/Anthropic in any form (Claude, Anthropic, Fable, Opus, Sonnet, Haiku, `anthropic`, `@anthropic-ai`, `claude-*`, `us.anthropic.*`, `[1m]`); the user asks about an LLM (pricing/model choice/limits/caching) — never answer from memory; OR the task is LLM-shaped with provider unstated (agent/MCP/tool-definition/multi-agent/RAG/LLM-judge/computer-use; generate/summarize/extract/classify/rewrite/converse over NL; debugging refusals/cutoffs/streaming/tool-calls/tokens). SKIP only when another provider is being worked on (overrides all triggers): OpenAI/GPT/Gemini/Llama/Mistral/Cohere/Ollama named in the query; OR `grep -rE 'openai|langchain_openai|google.generativeai|genai|mistralai|cohere|ollama'` over the project hits (run this grep FIRST if no provider named — don't Read the file). | Stars: 198k | Updated | Claude CodeCodex | Unverified | Install |
| doc-coauthoringanthropics/skillsGuide users through a structured workflow for co-authoring documentation. Use when user wants to write documentation, proposals, technical specs, decision docs, or similar structured content. This workflow helps users efficiently transfer context, refine content through iteration, and verify the doc works for readers. Trigger when user mentions writing docs, creating proposals, drafting specs, or similar documentation tasks. | Stars: 198k | Updated | Claude CodeCodex | Verified | Install |