video-product-pipeline

Orchestrates trend research, spec approval, generation, and auditing for short-form video production.

1|Updated Aug 7, 2026
One-click install
npx skills add https://github.com/Deepak-ai-93/deepak-skill --skill video-product-pipeline-deepak-ai-93
Or copy as Structured Prompt for Agent▼
Please help me install this Agent Skill.
Skill: video-product-pipeline
Source: https://github.com/Deepak-ai-93/deepak-skill/tree/main/skills/video-product-pipeline
Command: npx skills add https://github.com/Deepak-ai-93/deepak-skill --skill video-product-pipeline-deepak-ai-93

SYSTEM DOCUMENTATION & REQUIREMENTS

💡 This Skill requires playwright, and includes scripts (resource) components.

What problem does it solve? Short-form video requests often go straight from a vague prompt to a low-quality render with no research, no review, and no quality control. This Skill enforces a six-stage pipeline so every reel, TikTok, or Short is backed by live trend signals, a scored angle, an approved spec, and an automated audit before delivery. ## Core Features & Use Cases - Trend research and angle scoring: trend-hunt.mjs pulls Reddit top-of-day posts and Google Trends RSS without API keys, then a viral scorecard rates brainstormed angles out of 35. - Approval-gated spec: writes a complete video-product.md (beat sheet, safe-zone map, voice, decisions) and blocks all generation until the user explicitly approves. - Automated composition audit: audit-composition.mjs captures per-beat keyframes in headless Chrome, checks the 9:16 safe zone, text overlap, word caps, and GSAP timeline determinism, then hands off to an auditor subagent for sign-off. - Use Case: A user types "maek a vid about money" — the agent hunts trending finance topics, scores five angles, presents a spec for approval, delegates rendering to text-motion-reels and voice-sfx-audio, audits the frames, and delivers the MP4 with captions and reports. ## Quick Start Ask the agent to make a 15-second text-only reel about a topic and let it run the trend research, spec approval, generation, and audit pipeline.

Frequently Asked Questions about video-product-pipeline

High-intent search queries and answers about installing and using this skill.

FAQPage Schema
How do I make a viral short-form video with an AI agent?▼

Run the six-stage pipeline: hunt trending signals with trend-hunt.mjs, brainstorm and score at least five angles with the viral scorecard, write a video-product.md spec, get user approval, then delegate generation to text-motion-reels or video-asset-reels and audit the result before delivery.

How to audit an HTML video composition for safe zones and text overlap?▼

Run audit-composition.mjs with the --html flag and duration. It loads the composition in headless Chrome via Playwright, captures one keyframe per beat, measures every text element against the 9:16 safe zone (x 8-92%, y 15-85%), and writes an audit-report.md with PASS, WARN, or FAIL verdicts.

Does trend-hunt.mjs require API keys for Reddit or Google Trends?▼

No API keys are needed. The script fetches Reddit top-of-day posts via the public JSON endpoint and Google Trends through the Trending Now RSS feed, with failure-tolerant fallbacks if a source is unreachable.

What are the prerequisites for running the video pipeline scripts?▼

You need Node.js 18 or later, Google Chrome, FFmpeg, and Playwright installed as a dev dependency. Voice generation additionally requires Kokoro, and the pipeline orchestrates the text-motion-reels, video-asset-reels, and voice-sfx-audio skills.

Why does the pipeline block generation before user approval?▼

The approval gate exists because the common failure mode is prompt in, junk out. No HTML, rendering, or audio work begins until the user explicitly approves the video-product.md spec, which records the beat sheet, hook, safe-zone map, and all decisions.

What causes an audit FAIL in audit-composition.mjs?▼

FAILs are triggered by off-screen or clipped text, beats exceeding 10 words, Math.random or SMIL animations in the HTML, audio tags inside the composition, or a missing paused GSAP timeline. Any FAIL requires fixing, re-rendering, and re-auditing before delivery.