Video Workflow

End-to-end product video production through AI agent skills. Nine specialized skills, one orchestrator. Give your agent a project directory and a product URL — it handles everything from competitor research to final render.

The Pipeline

Phase 0: CONTEXT ──── Phase 1: STORYBOARD ──── Phase 2: PRODUCTION ──── Phase 3: VISUAL QA
  │                     │                        │                        │
  ├ product-context     ├ vsl-storyboard-writer  ├ recording-checklist    ├ Brand audit
  ├ searching-videos    ├ storyboard (6-frame)   ├ voiceover-tts         ├ Animation check
  ├ getting-videos      ├ audio-director plan    ├ audio-director asm    ├ Storyboard match
  └ analyzing-videos    └ seedance-prompts       └ Export + subtitles    └ User review

Phase 0 → 3 are documented across two pages: → Research Skills (Phase 0) — product-context, searching, getting, analyzing
→ Production Skills (Phase 2) — voiceover, audio, recording, Seedance

---

iopho-video-director

The master orchestrator. Call this first — it routes to all other skills automatically.

/iopho-video-director ./exp/my-project                    # new project
/iopho-video-director ./exp/my-project --mode continue    # resume from last checkpoint
/iopho-video-director ./exp/my-project --phase 2          # jump to specific phase

Modes

new (default) — Start fresh. Asks 5 onboarding questions, then runs Phase 0:

1. What product? — Name + URL
2. What kind of video? — Demo / Explainer / Launch / Teaser / Feature / Comparison
3. How long? — 15s / 30s / 60s / 90s / 120s+
4. Where will it live? — YouTube / Product Hunt / TikTok / App Store / B站 / XHS / LinkedIn
5. Any references? — Video URLs you like (fed into Phase 0 research)

Creates project-plan.md then begins Phase 0.

continue — Reads project-plan.md, detects last completed phase, resumes:

no project-plan.md    → switches to "new"
context.md missing    → Phase 0
storyboard.md missing → Phase 1
no out/video.mp4      → Phase 2
no QA pass noted      → Phase 3

jump --phase N — Jumps to a specific phase with prerequisite checks:

Phase 1 requires: context.md
Phase 2 requires: context.md + storyboard.md + audio-plan.md
Phase 3 requires: out/video.mp4

Phase checkpoints

Phase	Skills called	Key outputs	Checkpoint
0 Context	product-context → searching → getting → analyzing	context.md, research/storyboards/*.storyboard.md, pattern-analysis.md	User approves context summary
1 Storyboard	vsl-storyboard-writer → audio-director plan	storyboard.md, vo-script.md, audio-plan.md	User reviews storyboard — changes are cheapest here
2 Production	recording-checklist → voiceover-tts → audio-director assemble → seedance-prompts	out/video.mp4, audio/master-audio.mp3, *.srt	User watches final render
3 Visual QA	(inline — no separate skill)	QA pass	Brand + animation + platform checks

Phase 3 QA checklist

Visual QA runs inline without a separate skill. The director checks:

Brand compliance — colors, fonts, logo usage, tone match context.md
Animation restraint — no excessive zoom/pan, one focal animation per scene, no distracting transitions
Storyboard conformance — each scene matches storyboard intent, VO lines synced to visuals
Platform checks:

Platform	Check
YouTube	Thumbnail-worthy first frame? End screen space?
Product Hunt	Works without sound? Auto-play friendly?
TikTok / Reels	Vertical crop preserves key content?
App Store	Within 30s limit? Shows real app UI?
LinkedIn	Professional tone? Subtitles visible?

Project directory structure

After a full pipeline run:

{project-dir}/
├── project-plan.md          ← phase tracker + decision log
├── context.md               ← product/brand/audience context
├── storyboard.md            ← scene-by-scene breakdown
├── vo-script.md             ← VO lines + timecodes
├── audio-plan.md            ← BGM + VO + SFX strategy
├── recording-checklist.md   ← shot list for screen recording
├── research/
│   ├── storyboards/         ← reference .storyboard.md files
│   ├── downloads/           ← downloaded reference videos
│   └── pattern-analysis.md
├── public/
│   ├── videos/              ← raw screen recordings
│   ├── voiceover/           ← VO segments + master-vo.mp3
│   └── audio/               ← BGM + master-audio.mp3
└── out/
    ├── video.mp4            ← final render
    ├── video-vertical.mp4   ← 9:16 cut (if needed)
    ├── video.srt            ← subtitles
    └── video.vtt

---

Install

# Install all video skills at once
npx skills add iopho-team/iopho-skills --skill iopho-video-director
npx skills add iopho-team/iopho-skills --skill iopho-product-context
npx skills add iopho-team/iopho-skills --skill iopho-searching-videos
npx skills add iopho-team/iopho-skills --skill iopho-getting-videos
npx skills add iopho-team/iopho-skills --skill iopho-analyzing-videos
npx skills add iopho-team/iopho-skills --skill iopho-voiceover-tts
npx skills add iopho-team/iopho-skills --skill iopho-audio-director
npx skills add iopho-team/iopho-skills --skill iopho-recording-checklist
npx skills add iopho-team/iopho-skills --skill iopho-seedance-prompts

# Or all iopho skills at once
npx skills add iopho-team/iopho-skills --all -y

Global install (available across all projects):

npx skills add iopho-team/iopho-skills --all -g -y

Skill	Phase	Purpose
iopho-video-director	All	Master orchestrator
iopho-product-context	0	Project intake → context.md
iopho-searching-videos	0	Cross-platform video search
iopho-getting-videos	0	Download video/audio/subtitles
iopho-analyzing-videos	0	Reverse-engineer → .storyboard.md
iopho-voiceover-tts	2	Multi-engine TTS + assembly
iopho-audio-director	1–2	BGM + VO + SFX planning and mixing
iopho-recording-checklist	2	Screen recording shot list
iopho-seedance-prompts	1–2	Seedance 2.0 AI video generation

Source: github.com/iopho-team/iopho-skills