shotkit
Pre-production for founder-led video at scale. Brief becomes storyboard, shot specs, and per-generator prompts in minutes.
Video production has a pre-production bottleneck. Brief becomes vague PDF, PDF becomes wasted shoot day, shoot day becomes 3-week edit. shotkit removes the bottleneck by turning the brief into structured storyboard files and per-generator prompts in minutes.
The Problem
Most founder-led video dies in pre-production. The brief is vague, the storyboard never gets written, the prompts get re-invented for every shot. By the time generation starts, half the budget is gone.
What It Does
- 01Brief becomes a structured storyboard with shot specs in under 10 minutes
- 02Per-generator prompts auto-generated for Veo, Kling, Flux, Runway, and Sora
- 03Versioned audit trail so every shot has provenance
What shotkit actually does
shotkit is four composable Claude Code skills that handle the work between "I have an idea for a video" and "the generators are running."
The four skills:
- storyboard-architect — Takes a brief in plain English. Outputs a structured
storyboard.mdandshots.jsonwith one entry per shot. - visual-prompt-forge — Takes the shots and generates seven generator-specific prompt files. One prompt for Veo, one for Kling, one for Flux, one for Runway, and so on. Each prompt is tuned to that model's strengths.
- visual-asset-critic — A QA loop that scores generated images against the original brief. Flags shots that need re-rolling.
- storyboard-html-preview — Renders the whole storyboard as a single shareable HTML file. No build step. Email it to your client.
Together, these four skills compress a typical 2-day pre-production cycle into about 30 minutes of focused work.
The math
A typical video agency charges $3,000–$10,000 just for pre-production on a 60-second piece. shotkit doesn't replace the thinking — you still need a brief, you still need a creative idea — but it eliminates the grinding. The shot list, the prompt engineering for six different generators, the documentation, the storyboard PDF — that's the work shotkit removes.
We use shotkit internally to produce video for every founder-led brand we operate. Every video we render — hundreds per client — starts with a shotkit-generated storyboard.
How it composes with the other skills
shotkit reads the brand voice profile generated by media-tsunami. If a brand-config.json exists for the brand, shotkit automatically locks the visual style, allowed words, forbidden phrases, and CTA structure into every shot's prompt. You don't have to remember the brand rules — they're enforced at generation time.
When combined with slopfiles, shotkit handles general video production while slopfiles specializes in serialized investigative content (like The Slop Files). Both feed the same downstream renderers.
Install
git clone https://github.com/whystrohm/shotkit.git
cd shotkit
./install.sh
The installer copies the four skill files into ~/.claude/skills/. Restart Claude Code and you'll see four new commands: /storyboard-architect, /visual-prompt-forge, /visual-asset-critic, /storyboard-html-preview.
Full docs and source on GitHub →
What you get
After running shotkit on a single brief:
my-video-project/
├── brief.md ← your input
├── storyboard.md ← human-readable shot breakdown
├── shots.json ← structured shot data
├── prompts/
│ ├── veo3.md
│ ├── kling.md
│ ├── flux-pro.md
│ ├── runway-gen3.md
│ ├── luma.md
│ ├── sora.md
│ └── elevenlabs.md ← for narration
├── preview.html ← shareable preview
└── audit-log.json ← versioned history
That's what comes out of a single /storyboard-architect run. Drop the prompts into fal.ai, get your first-frame/last-frame stills back, run them through your animator of choice. The Slop Files Episode 1 was produced exactly this way.
How It Composes
shotkit sits at the production layer. It reads brand-config.json from media-tsunami to lock visual style and voice into the storyboard. Combined with slopfiles for serialized output, it's the pipeline behind The Slop Files.
Related Skills
media-tsunami
The empirical layer. Extracts brand voice as executable code — cadence, vocabulary, forbidden words, exemplar sentences — serialized as a CLAUDE.md any LLM can load.
Install →slopfiles
The skill pack we use to produce The Slop Files series. Brief becomes 6 first-frame/last-frame video prompts in minutes.
Install →