Communitygithub.com

Gavin152/promo-clip

Guided interview for a promo-clip video — writes brief.md, the theme and the output settings.

Was ist promo-clip?

promo-clip is a Claude Code agent skill that guided interview for a promo-clip video — writes brief.md, the theme and the output settings.

Funktioniert mit✓Claude Code~Codex CLI~Cursor
npx skills add https://github.com/Gavin152/promo-clip/tree/HEAD/skills/discover

In Ihrer bevorzugten KI fragen

Öffnet einen neuen Chat, in dem dieser Agent-Skill bereits geladen ist.

Dokumentation

/promo-clip:discover

Read ${CLAUDE_PLUGIN_ROOT}/skills/promo-clip/protocol.md first. Then stage discover start.

1. Pre-fill

Before asking anything, collect what is already known so the interview confirms instead of asks:

  • an existing V/brief.md (a re-run: show each answer and ask only what should change);
  • promo.json / brief.md of a parent or sibling video folder (same product, earlier clip);
  • brand material at a folder or URL the user names: logo (SVG/PNG), colours, fonts, app screenshots, website tagline. Copy files into V/assets/ref/ (style references) or V/assets/img/ (logo, screenshots, badges — used as-is).

2. Interview

One question at a time, multiple choice with AskUserQuestion wherever the answer has a natural set (offer pre-filled answers as the first, recommended option). Topics, in order:

  1. Project — name, one-line description, website, brand material.
  2. Goal and audience — who it is for, what they should do afterwards.
  3. Key messages — 3–5 messages, plus the call to action.
  4. Length — 15 s, 30 s, 60 s, or custom.
  5. Output — formats (any of 16:9, 9:16, 1:1), the primary format, sizes (720p, 1080p, 1440p, 4K or WxH), fps (24, 25, 30, 60).
  6. Tone and story shape — problem → reveal → how it works → CTA; feature tour; testimonial-style.
  7. Voice — AI voice (language, gender, energy; offer 2–3 voices from V/providers.json tts.voices) or the user's own recording (path).
  8. Music — AI music (genre, mood, instruments, where it drops and lifts) or the user's own track (path).
  9. Visuals — main characters, settings, must-have screens (screenshots), own photos/assets, style references.
  10. End card — logo (text or image), QR target URL, store badges, website, tagline.
  11. Mode and budget — gated or auto; a cost limit in USD (a typical 30 s clip spends $1–3 on AI assets).

Done when every topic has an answer (or an explicit "not needed").

3. Write

  • V/brief.md — one ## <topic> section per topic above, in that order, answers as short bullets. The ## Music section holds a style-only description (genre, mood, instruments, energy curve) — the music stage feeds it to the generator as is.
  • V/theme.json — name, palette (brand colours mapped onto the engine's palette roles), logo, endCard, brandWords (lower-case brand words, for voice QA), and fontFiles for a brand font the user supplied. Palette roles and the theme fields: ${CLAUDE_PLUGIN_ROOT}/skills/promo-craft/theme.md.
  • Output, mode, budget through the CLI: output --formats … --primary … --sizes … --fps …, mode gated|auto, budget <usd>.
  • A QR target: art qr <url>; a logo image or screenshots in assets/img/: art manifest.

4. Gate

Show a one-page summary of the brief (product, audience, messages + CTA, length, formats/sizes/fps, tone, voice, music, visuals, end card, mode, budget). Gate per the protocol; changes loop back to 2 for the affected topics only. Then stage discover done.

Individual skills in this repo

This repo contains 8 individual skills — each has its own dedicated page.

Verwandte Skills