gistfile1.txt

---
name: video-publish-workflow
description: Complete YouTube video publishing workflow that chains thumbnail generation, upload, transcript extraction, and content repurposing. Use when user says "publish video", "full video workflow", "upload and repurpose", or provides a video file with title for end-to-end YouTube publishing with content distribution.
---

# Video Publish Workflow

Orchestrate the complete video publishing pipeline by chaining existing skills.

## Required Inputs

| Input | Required | Description |
|-------|----------|-------------|
| Video file | Yes | Path to the video file |
| Title | Yes | Video title |
| Transcript | No | If provided, skip transcript extraction step |
| Description hints | No | Key points to emphasize in description |

## Workspace Setup

**Before starting the workflow**, create a dedicated workspace directory:

```
~/Videos/workflows/{YYYY-MM-DD}-{title-slug}/
```

Example: `~/Videos/workflows/2024-12-05-how-to-build-landing-page/`

All workflow outputs go in this directory:

```
workspace/
├── thumbnail.png          # Generated thumbnail (original)
├── thumbnail.jpg          # Compressed thumbnail (uploaded)
├── transcript.txt         # Full transcript
├── description.txt        # YouTube description
├── blog.md               # Blog post draft
├── linkedin.md           # LinkedIn post draft
└── tweet.md              # Twitter thread draft
```

**When invoking other skills, save all outputs to the workspace directory** instead of default /tmp locations. This preserves outputs from previous workflow runs.

## Workflow Decision Tree

```
Has transcript provided?
├─ Yes → Use provided transcript
└─ No → Extract locally with local-transcribe (fast, ~10 seconds)
        └─ Fallback: youtube-transcript if local fails
```

## Workflow A: Transcript Provided

Execute sequentially:

1. **Generate thumbnail** → `youtube-thumbnail-generator` skill
   - Input: video title
   - Output: thumbnail image (PNG)
   - **Compress to JPEG <2MB** before upload (YouTube limit)
   ```python
   from PIL import Image
   img = Image.open('thumbnail.png')
   img.convert('RGB').save('thumbnail.jpg', 'JPEG', quality=85, optimize=True)
   ```

2. **Generate description** from transcript
   - Extract key points, timestamps if video >5 min
   - Include hooks and CTAs
   - Keep under 5000 characters

3. **Upload to YouTube** → `youtube-upload` skill
   - **Use venv**: `~/.claude/skills/youtube-upload/venv/bin/python`
   - Input: video file, title, description, thumbnail (compressed .jpg)
   - Privacy: start with `unlisted` unless user specifies
   - Output: video URL, video ID

4. **Repurpose content** → `content-repurpose` skill
   - Input: transcript, video URL
   - Output: blog post draft, Twitter thread, LinkedIn post

5. **Present results** to user:
   - Video URL
   - Content drafts for review
   - Suggested next steps

## Workflow B: No Transcript (Recommended Flow)

Execute sequentially:

1. **Extract transcript locally** → `local-transcribe` skill
   - **Use venv**: `~/.claude/skills/local-transcribe/venv/bin/python scripts/transcribe.py`
   - Input: video file path
   - **Use `--timestamps` flag** for chapter generation in description
   - Output: timestamped transcript (~10 seconds for most videos)
   - Save to: `workspace/transcript.txt`
   - Fallback: If local transcription fails, proceed with upload and use `youtube-transcript` after

2. **Generate thumbnail** → `youtube-thumbnail-generator` skill
   - Input: video title
   - Output: thumbnail image (PNG)
   - **Compress to JPEG <2MB** before upload (YouTube limit)

3. **Generate rich description** from transcript
   - Include timestamps, key points, CTAs
   - Save to: `workspace/description.txt`

4. **Upload to YouTube** → `youtube-upload` skill
   - **Use venv**: `~/.claude/skills/youtube-upload/venv/bin/python`
   - Input: video file, title, full description, thumbnail (compressed .jpg)
   - Privacy: `unlisted` unless user specifies
   - Output: video URL, video ID

5. **Repurpose content** → `content-repurpose` skill
   - Input: transcript, video URL
   - Output: blog post draft, Twitter thread, LinkedIn post

6. **Present results** to user

## Description Generation Guidelines

When generating video descriptions from transcript:

```
[Hook - 1-2 compelling sentences]

[Main content summary - 2-3 paragraphs]

⏱️ Timestamps:
0:00 - Intro
[Extract key sections from transcript]

🔗 Links:
[Relevant links if mentioned]

#hashtags #relevant #totopic
```

Keep descriptions:
- Under 5000 characters
- Front-loaded with important info (first 150 chars show in search)
- Include 3-5 relevant hashtags at end

## Example Usage

**User says:** "Publish this video: /path/to/video.mp4 with title 'How to Build a Landing Page in 10 Minutes'"

**Execute:**
1. Create workspace: `~/Videos/workflows/2024-12-05-how-to-build-landing-page/`
2. Extract transcript locally (~10 sec) → save to `workspace/transcript.txt`
3. Generate thumbnail → save to `workspace/thumbnail.png`, compress to `workspace/thumbnail.jpg`
4. Generate full description with timestamps → save to `workspace/description.txt`
5. Upload video as unlisted with full description and thumbnail
6. Repurpose into blog/Twitter/LinkedIn → save to `workspace/*.md`
7. Present all outputs to user with workspace path

## Skill Dependencies

This workflow requires these skills to be available:

| Skill | Purpose |
|-------|---------|
| `local-transcribe` | Extract transcript locally (primary, fast) |
| `youtube-thumbnail-generator` | Generate video thumbnail |
| `youtube-upload` | Upload video, set thumbnail, update metadata |
| `youtube-transcript` | Extract transcript from YouTube (fallback only) |
| `content-repurpose` | Create blog/Twitter/LinkedIn content |

## Error Handling

- **Local transcription fails**: Fall back to `youtube-transcript` skill after upload (will need to wait 5-10 min for YouTube processing)
- **Thumbnail generation fails**: Proceed with upload, user can set thumbnail manually
- **Upload fails**: Report error, do not proceed
- **YouTube transcript unavailable**: Wait and retry, or ask user to provide transcript
- **Repurpose fails**: Still report successful upload, note repurpose issue

## Post-Workflow

After workflow completes, present a summary:

```
Workflow Complete!

Video: https://youtu.be/{video_id}
Status: Unlisted (change to public when ready)

Workspace: ~/Videos/workflows/{date}-{slug}/
├── thumbnail.jpg    ✓
├── transcript.txt   ✓
├── description.txt  ✓
├── blog.md         ✓
├── linkedin.md     ✓
└── tweet.md        ✓

Next steps:
1. Review content drafts in workspace
2. Post to blog/Twitter/LinkedIn manually
3. Change video to public when ready
```

All outputs are preserved in the workspace for future reference or re-use.