All skills
hoodini avatar

/yuv-video-director

@f412198

Yuval's all-in-one AI video pipeline. Turns an idea/script into a finished, on-brand MP4 by orchestrating HyperFrames (HTML→deterministic video render), Lottie (branded motion graphics), ManimCE (math / neural-network / concept animations), and a transcribe→approve caption flow — all wrapped in the YUV.AI Neon Phoenix brand via a frame.md. Use whenever Yuval wants to make, edit, or explain something as a video: promo, explainer, launch, social reel, "make a video about X", "explain X as a video", "neural network animation", "turn this into a video", captioned tutorial, 16:9 or 9:16. Triggers: video, explainer, promo, reel, manim, lottie, hyperframes, animation, "make a video", "explain ... as a video", מצגת וידאו, סרטון, הסבר וידאו. Routes each beat to the right engine, wraps in brand, self-verifies, and renders.

Use this Skill: https://skilld.dev/gh/hoodini/ai-agents-skills/yuv-video-director

This session only. Nothing lands on disk.

referencesgates.md

≈476 tokens on demand. Your agent reads this file only when SKILL.md points to it.

Gates — self-verify & self-heal loop

Run after every authoring change. Fix → re-run until clean. This is what makes the pipeline agentic.

Per-stage gates

  1. Manim (before compositing): test-render -ql first to catch API errors fast; verify the MP4 exists + duration matches the beat; spot-check 2 frames. Then -qh final.
  2. Lottie: after wiring, screenshot a mid-intro frame and confirm it actually renders shapes (the Skottie↔lottie-web trap). Blank → fix the JSON (group-wrap, keyframe i/o handles).
  3. Composition lint: npx hyperframes lint → 0 errors. Common fixes:
    • overlapping_gsap_tweens → add overwrite: "auto"
    • font_family_without_font_face → Anton isn't auto-resolved; download the woff2 + add @font-face (Inter/JetBrains Mono auto-resolve). Use the latin subset woff2.
    • remove Google Fonts <link> (fails in sandboxed render) — go local for Anton.
  4. Validate: npx hyperframes validate → 0 console errors (proves Lottie loaded + video resolved) + WCAG AA on all text. Failing contrast → adjust the failing color within the palette (brighten on ink), or add a radial ink scrim behind text. Judge badge/decorative warnings.
  5. Render → spot-check 5 frames across the timeline (cover, each engine beat, closing). Confirm each engine is actually visible (field, Lottie, Manim clip, captions) and nothing overlaps/clips.

Determinism check (for any live canvas/adapter)

Seeking the same t twice must produce the same frame. No Date.now()/Math.random() (seed a mulberry32), no setTimeout, no .play(). The neural-net field is driven by a GSAP proxy onUpdate — that's seekable.

Don't silently cap

If you drop a beat, shorten the video, or skip an engine (e.g. Manim not installed), say so — don't present a reduced video as if it were the full ask.

Source: SKILL.md on GitHub

1 warning3mo3 checks · Risk SAFE
  • Gen Agent Trust Hub3mo

    The yuv-video-director skill is a comprehensive and well-architected video production pipeline for AI agents. It orchestrates HyperFrames, Lottie, and ManimCE to create professional-grade video content with synchronized captions and branded motion graphics. The skill follows standard practices for asset management, using helper scripts to fetch legitimate resources like fonts and libraries from trusted sources. No malicious patterns, obfuscation, or data exfiltration risks were identified during the analysis.

  • Socket3mo

    No alerts

  • Snyk3mo

    Risk: MEDIUM · 1 issue

Signed by skilld at f412198. This ties the file your Agent reads to that commit on GitHub. It does not review the instructions.

Last checked against GitHub 2 days ago.

Steadyupdated 4 months ago

README badge

README badge for hoodini/ai-agents-skills/yuv-video-director