# Explainer

> Build a bespoke explainer as an interactive HTML page, a narrated 3Blue1Brown-style video, or an ASD-STE100 technical description page. Invocation-only. Use for /explainer, $explainer, or an explicit request to use the explainer skill. Not for quick inline charts, ELI5 chat answers, product demo videos, or /explain lessons.

- Skill: `mattdweigand-sketch/explainer` (Agent Skill, multi-file: 32 files)
- Install (CLI): `npx get-skill add mattdweigand-sketch/explainer`
- Raw SKILL.md: https://api.skillmd.com/api/skills/mattdweigand-sketch/explainer/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Web & Frontend
- Author: mattdweigand-sketch (https://skillmd.com/u/mattdweigand-sketch)
- Updated: 2026-11-16
- Page: https://skillmd.com/skills/mattdweigand-sketch/explainer

---


# Explainer

Build a large, custom, discardable artifact whose only job is to make one topic understood.

## Pick the mode

Ask every time, before any research or drafting, even when the request names a format. Use one single-select question with **Web page**, **Video**, and **Technical description** as the options, then wait for the answer. If the request names a format, list that option first.

- **Web page.** Named by "in HTML", "web page", or "interactive". Read `references/web.md`.
- **Video.** Named by "video", "3b1b", "animated video", or "narrated". Read `references/video.md`.
- **Technical description.** Named by "STE", "ASD-STE100", "Simplified Technical English", "technical description", or "description and operation". Read `references/technical.md`.

## Rules for all modes

1. Pin the one sentence the viewer should leave with and the audience level. Infer both from the request. Ask only when the audience would change the content.
2. Establish the facts before drafting. Use supplied material first. Verify facts that change over time. Separate facts, interpretations, unknowns, and examples. Cite as the mode file says. Video narration cannot carry links, so list its sources in the handoff.
3. Label illustrative data as illustrative.
4. Order the ideas: a question or puzzle, one concrete example, the mechanism, the general rule, a one-line recap. Technical description uses the section order in `references/technical.md` instead.
5. Show, then say. Every beat has one visual that carries the idea. Text and narration support the visual, not the reverse. Technical description uses the figure count in `references/technical.md` instead.
6. Read `references/writing.md` before writing narration or page text. Define each term on first use. Technical description applies the full standard in `references/technical.md` instead.
7. Draw diagrams, labels, and equations with code. Never use image or video generation for them, because generated media cannot render exact labels.
8. Use the shared dark theme. `templates/page.html` and `templates/explainer_scene.py` hold the same dark palette, role colors, and PPLX type from `assets/fonts/`. The page adds a light variant behind a toggle. After editing either file, run `python scripts/test_theme_parity.py` from this skill directory. Technical description takes its theme from `templates/page.html` when `scripts/build_technical.py` runs.

## Deliver

- Report the actual result of the mode's Check list and its limits. Never infer browser verification from code inspection.
- List any claim you could not verify.
- Never call output STE compliant, certified, or ASD approved. In Technical description mode, say that a script examined the page and that ASD did not.
- When the execution target is a RemoteBox, the website deploy tool is unavailable there (seen 2026-10-12). Copy the built page to the Cloud sandbox with the file-copy tool, switch to Cloud, deploy, then switch back.

