Short Drama Publisher

Automated short drama video publisher. Downloads drama content from MoboBoost, uses AI to identify highlight moments, clips 15-second vertical videos with te...

Security Scan

The SKILL.md describes a downloader/publisher that requires cookies and scripts, but the skill bundle contains no code, no homepage/source, and metadata mismatches the instructions — proceed only after verifying the missing code and carefully protecting account cookies.

Purpose & Capability

The described purpose (download from MoboBoost, clip, publish to Facebook) would reasonably need ffmpeg, Python, Playwright, and account cookies — which the SKILL.md documents. However, the package contains no scripts or code (despite referencing scripts/ and a full project layout), and registry metadata lists no required env vars while the SKILL.md declares MOBOBOOST_COOKIES and FACEBOOK_COOKIES. The absence of actual code or a source/homepage for the implementation is a major incoherence.

Instruction Scope

Runtime instructions tell the user/agent to run specific Python scripts, export browser cookies, and cron the workflow. Those actions are consistent with the stated task, but because the scripts are not included, the instructions cannot be executed as-is and the agent would have no vetted code to run. The instructions also rely on cookie-based auth (sensitive), and do not provide any guidance to safely handle or limit cookie scope.

Install Mechanism

There is no install spec (instruction-only) which limits automatic risk from bundle installation. The SKILL.md suggests installing ffmpeg, pip packages, and Playwright—standard for a local workflow. Because no archive downloads or third-party install scripts are declared in the bundle, there is lower install-time risk, but the user would still need to manually install packages and fetch or implement the missing scripts.

Credentials

Requested credentials (MoboBoost and Facebook login cookies) are directly related to the publishing/downloading task. That said, cookies are extremely sensitive (they can grant account access). The SKILL.md asks to store cookies in local JSON files without recommending secure storage or minimization. Also metadata/registry claims no required env vars while SKILL.md documents cookie env/config keys — an inconsistency to resolve.

Persistence & Privilege

The skill is not marked always:true and does not request persistent elevated privileges. Cron scheduling is user-provided example only. Autonomous invocation is allowed by platform defaults but is not combined with any special persistent privileges in the bundle.

What to consider before installing

This package is instruction-only: it documents scripts and a full project layout (moboboost_downloader.py, highlight_detector.py, facebook_publisher.py, etc.) but does not include any of those files or a source/homepage. Do not try to run anything you copy from the SKILL.md as-is — there is nothing in the bundle to run. Before installing or following these instructions: 1) obtain and inspect the actual implementation code from a trusted source; 2) verify licensing/rights to download and repost MoboBoost content; 3) avoid exporting/storing account cookies in plaintext unless you understand the risk — cookies can enable account takeover; prefer OAuth or official APIs where possible; 4) confirm the SKILL.md metadata (env vars, requirements) matches the real code; and 5) if you decide to run third-party scripts, review them line-by-line for exfiltration or unexpected network calls. The mismatch between declared metadata and the SKILL.md plus the lack of source code is the primary reason this is flagged suspicious.

Short Drama Publisher (短剧自动化发布)

Automated short drama promotion video workflow, inspired by Strawberry TV model:

1. 📥 Download drama content from MoboBoost
2. 🎯 AI-powered highlight detection (scene changes, audio peaks, subtitle emotion)
3. ✂️ Clip 15-second vertical videos with white English title overlay
4. 📤 Auto-publish to Facebook

---

Prerequisites

System Dependencies

# macOS
brew install ffmpeg

# Python dependencies
pip install playwright opencv-python librosa numpy pyyaml
playwright install chromium

Credentials Setup

1. MoboBoost Cookies

   • Login to https://ckoc.cdreader.com
   • Export cookies using browser extension (e.g., "EditThisCookie")
   • Save as config/moboboost_cookies.json

2. Facebook Cookies

   • Login to Facebook
   • Export cookies using browser extension
   • Save as config/facebook_cookies.json

---

Usage

Full Automated Workflow

python scripts/daily_workflow.py

Individual Modules

Download content:

python scripts/moboboost_downloader.py --drama-code 613815

Detect highlights:

python scripts/highlight_detector.py --input data/downloads/video.mp4

Clip video:

python scripts/video_editor.py --input video.mp4 --start 01:23 --title "Drama Name"

Publish to Facebook:

python scripts/facebook_publisher.py --video data/outputs/clip.mp4 --drama-code 613815 --drama-name "DramaName"

Daily Cron Job

# Run daily at 9am
0 9 * * * cd /path/to/short-drama-publisher && python scripts/daily_workflow.py >> logs/cron.log 2>&1

---

Configuration

settings.yaml

# Video settings
video:
  duration: 15          # Clip duration (seconds)
  width: 1080           # Width
  height: 1920          # Height (9:16 vertical)

# Text overlay settings
text_overlay:
  font: "Arial-Bold"
  size_ratio: 0.05      # Font size as ratio of video width
  color: "#FFFFFF"
  border_color: "#000000"
  border_width: 2
  position_y: 0.75      # Vertical position (ratio from top)

# AI highlight detection weights
highlight_weights:
  scene_change: 0.30
  audio_peak: 0.25
  subtitle_emotion: 0.25
  motion_intensity: 0.20

# Publishing settings
publishing:
  videos_per_day: 3     # Number of videos per day
  interval_minutes: 120 # Interval between posts (minutes)

---

Directory Structure

short-drama-publisher/
├── SKILL.md                    # This file
├── config/
│   ├── settings.yaml           # Configuration
│   ├── moboboost_cookies.json  # MoboBoost credentials
│   └── facebook_cookies.json   # Facebook credentials
├── scripts/
│   ├── moboboost_downloader.py # Content downloader
│   ├── highlight_detector.py   # AI highlight detection
│   ├── video_editor.py         # Video clipping
│   ├── facebook_publisher.py   # Facebook publisher
│   └── daily_workflow.py       # Main workflow
├── data/
│   ├── downloads/              # Raw downloaded content
│   ├── outputs/                # Clipped videos
│   └── history.json            # Publishing history
├── fonts/                      # Font files
└── logs/                       # Log files

---

AI Highlight Detection

The highlight detector uses multiple signals to find the most engaging moments:

Signal	Weight	Method
Scene Change	30%	OpenCV frame-by-frame difference analysis
Audio Peak	25%	Librosa audio amplitude analysis
Subtitle Emotion	25%	Text sentiment analysis on subtitles
Motion Intensity	20%	Optical flow magnitude calculation

Each frame gets a composite score, and the highest-scoring 15-second segment is selected.

---

Important Notes

[!WARNING]

• MoboBoost and Facebook websites may update, requiring script adjustments
• Recommend 1-3 videos per day to simulate organic posting rhythm
• Ensure you have rights to use MoboBoost content for promotion
• Cookie-based auth may expire; re-export periodically