One mechanism. A visual experiment.
Blueprint Lab turns a short English script into a vertical knowledge video. Use animated flows, discrete quantities and equations on an original navy drafting grid. Keep your sources, assumptions and editable scene data with the film.
Download and use
Download Blueprint Lab 0.1.1 (ZIP)
Extract the ZIP into your project’s .agents/skills/ folder. Open the project in Codex and ask:
Read .agents/skills/blueprint-explainer/SKILL.md. Make a narrated 36-second vertical video explaining why a queue grows. Use the bundled queue example, preserve its assumptions, and export to a new folder.
The package works from local files. An account is not needed. FeedSkills CLI can also install it from the public registry; the CLI is available as a downloadable package, not published on npm:
npx --yes --package=https://feedskills.com/downloads/feedskills-0.3.6.tgz feedskills install blueprint-explainer --agent codex
Inputs, outputs and local dependencies
- 4–6 scenes, 30–45 seconds total, one concept, English text, and source URLs with notes.
- Output: 1080×1920 / 30 fps MP4, cover JPG, SRT, publish copy, source notes and editable JSON + HTML.
- macOS, Node.js 22+, FFmpeg and FFprobe. The first render downloads HyperFrames 0.8.181 and a headless browser.
- Optional local Kokoro narration requires Python 3.10+, kokoro-onnx 0.6.1 and soundfile 0.14.0; first use downloads about 338 MB of model and voice data. There is no paid generation API. Agent-provider fees are separate.
- The bundled queue and cache inputs demonstrate different mechanisms. No narration is included unless you supply audio or run the local narration helper.
Run contract
node .agents/skills/blueprint-explainer/scripts/project.mjs doctor
node .agents/skills/blueprint-explainer/scripts/project.mjs create .agents/skills/blueprint-explainer/examples/queue.json --out ./queue-video
cd queue-video
npm run check
npm run preview
npm run render
Use a new project folder for each revision. Edit the JSON for content changes or the generated HTML for custom motion. Review the complete exported film for meaning, pacing, text and audio before publishing.
License and tested scope
Free and source available under the FeedSkills Source-Available License 1.0. You may publish and sell finished videos, including client work. Redistribution, resale and rebranding of the Skill files are not permitted. HyperFrames is a separate Apache-2.0 dependency; this package contains no bundled GSAP, fonts, stock media or model weights.
Tested on macOS with Codex. Fresh operating-system setup, other Agent hosts and real social-app safe zones remain unverified. This release makes no promise about views, customer adoption or support for every subject.