Contenu du pack

Un fichier SKILL.md (le guide pratique) + un modèle de départ. Aperçu :

Multi-Agent Capability Reel

When to use

• You are launching a product where several AI agents work together and a single screenshot cannot show the collaboration
• You need a short social video that demonstrates a capability — agents handing work to each other — rather than describing it
• Your launch announcement needs a visual that works muted, since most feed viewers watch with sound off
• You want a repeatable reel format: same structure, same pacing, different task each time
• You have screen recordings or demo clips of agents working and need them cut into a tight story

Workflow

1. Pick one task (input: product capabilities → output: reel task + beat list)
   Choose a single concrete task the agents complete together — one input, one finished output. Break it into 3–5 beats, one per agent handoff: agent A gathers, agent B drafts, agent C checks, and so on. If the task needs more than five beats to explain, it is two reels, not one. Write the beat list as plain sentences before touching any footage.
2. Script for muted viewing (input: beat list → output: caption script)
   Write captions first, not voiceover: every beat gets one short caption line (under 8 words) that states what is happening on screen. Captions carry the whole story — a viewer with sound off must understand the reel completely. Keep total runtime in the 30–60 second range; each beat gets roughly 6–10 seconds.
3. Assemble the footage (input: caption script + clips → output: rough cut)
   Cut each beat to its caption: show the handoff moment, not the waiting. Use screen recordings, product UI, or simple motion graphics — whatever shows the work changing hands. Cut on the action: the moment agent B picks up agent A's output is the moment to switch shots. Remove any clip where nothing visibly changes for more than two seconds.
4. Pace and caption (input: rough cut → output: captioned cut)
   Set the rhythm: quick cuts during handoffs, a one-beat pause on the final result. Burn captions into the video in large, high-contrast type, centered or lower-third, timed to the action — never auto-generated captions left unedited. Add a simple music bed or none at all; if music is used, it must sit under the captions, not compete with them. End on the finished output with a single call to action.
5. Check before posting (input: captioned cut → output: publish-ready video)
   Watch the reel three ways: muted on a phone, with sound, and at 2x speed. Verify every caption is spelled correctly, every agent's role matches what the product actually does, and the final output shown is a real output, not a mock-up presented as real. Export in the platform's preferred format and aspect ratio, and keep a captioned transcript with the post for accessibility.

Aperçu du modèle

Multi-Agent Capability Reel — Script & Cut Template

One task = one reel. Script the captions before touching any footage.

The task

Task (one input → one finished output): {task} Input: {what goes in} | Finished output: {what comes out} Runtime target: {30–60 seconds} | Aspect ratio: {vertical / square}

Beat list (3–5 handoffs, one per agent)

Cut rule: show the handoff moment, not the waiting. Remove any clip where nothing visibly changes for more than 2 seconds.

Caption script (one line per beat, under 8 words each)

Call to action (one, on the final frame): {CTA}

Pacing & captions

Le pack de compétence lui-même (SKILL.md + modèle) est rédigé en anglais.

Acheter le pack complet — $4.99 Pro — les 511 compétences