メニュー
Analyze YOUR video recordings before publishing. Evaluates voice (pace, pitch, volume), facial expressions (emotions, eye contact, smiles), and content (filler words, structure). Helps improve your Hinglish YouTube delivery over time.
Codex または Claude でインストール この Prompt をコピーして Codex、Claude、または他のアシスタントに貼り付けると、Skill ページを確認してインストールできます。
SOC 職業分類に基づく
Transforms research and knowledge base material into information-dense infographic copy—clear, scannable, authoritative, and reference-worthy. Use when converting clinical algorithms, decision trees, comparison data (drug vs drug, treatment options), trial results, myth vs truth content, patient preparation guides, checklists, or sequential process workflows into structured copy ready for single-image infographic formats. Supports multiple infographic types including hero stats, dense multi-section layouts, side-by-side comparisons, myth-busters, step-by-step processes, and patient checklists. Outputs structured copy with compressed noun-heavy phrasing, specific numbers, trial citations, and JSON-formatted handoff data for downstream infographic generation.
Manages all other skills by detecting capability gaps, proposing new skills, and keeping the system registry in sync. Use when the user encounters something the system cannot do, wants to log a capability gap, review the skill backlog, propose or build a new skill, analyze recurring unmet needs, or run sync pipelines to update context files (CLAUDE.md, GEMINI.md, AGENTS.md) after skill changes.
Template-driven medical infographics using the AntV Infographic framework. Use when creating structured infographics from clinical data, trial results, or patient education content using pre-built templates for hero stats, comparison layouts, process flows, myth-busters, and checklists. Outputs publication-ready HTML/PNG infographics with JACC/NEJM-grade design tokens.
Declarative, grammar-based charting using AntV G2 for complex medical visualizations. Use when creating forest plots, Kaplan-Meier curves, funnel plots, waterfall charts, or any statistical/clinical chart that requires grammar-of-graphics composition with medical-journal-grade styling and accessibility-validated color palettes.
Publication-grade design tokens and utilities for Nature/JACC/NEJM quality graphics. Use when generating charts, infographics, animations, or any visual content that requires medical-journal-grade color palettes, typography, accessibility-validated contrast ratios, and consistent branding. Provides Python token APIs, colorblind-safe palettes, G2 chart templates, AntV infographic templates, Vizzu animation presets, and Manim integration for animated medical explainers.
Content OS orchestrator - the master skill that produces ALL content types from one seed idea (forward mode) or splits long-form content into short-form pieces (backward mode). Invokes research, writing, quality review, and visual generation skills in a coordinated pipeline. Long-form content goes through full quality gates; short-form gets quick accuracy pass.
| name | video-delivery-coach |
| description | Analyze YOUR video recordings before publishing. Evaluates voice (pace, pitch, volume), facial expressions (emotions, eye contact, smiles), and content (filler words, structure). Helps improve your Hinglish YouTube delivery over time. |
Get better at video, video by video. This skill analyzes your recordings before you publish, identifying areas for improvement.
| Analysis Type | Metrics | Tool Used |
|---|---|---|
| Voice | Speech rate (WPM), pitch variation, volume consistency | Librosa + Whisper |
| Facial | Emotion timeline, eye contact frequency, smile frequency | OpenCV + DeepFace + Mediapipe |
| Content | Transcription, filler words, structure | Faster-Whisper + Claude |
| Overall | 5-dimension score (1-5 each, max 25) | Claude analysis |
| Dimension | Score 1 | Score 5 |
|---|---|---|
| Content & Organization | Disorganized, unclear | Logical, well-structured |
| Delivery & Vocal Quality | Monotone, many fillers | Clear, varied, engaging |
| Body Language & Eye Contact | No eye contact, stiff | Direct gaze, natural movement |
| Audience Engagement | Boring, loses attention | Captivating, maintains interest |
| Language & Clarity | Grammar issues, unclear | Clear, impactful, professional |
Total Score Interpretation:
Use this skill when you say:
"Analyze my video at /path/to/recording.mp4"
"Coach my delivery on the latest YouTube recording"
"What can I improve in this video?"
# Basic analysis
python scripts/analyze_video.py --video "/path/to/video.mp4"
# Full analysis with all features
python scripts/analyze_video.py --video "/path/to/video.mp4" --full
# Voice only (faster)
python scripts/analyze_video.py --video "/path/to/video.mp4" --voice-only
# Save report
python scripts/analyze_video.py --video "/path/to/video.mp4" --output ~/reports/
┌────────────────────────────────────────┐
│ VIDEO DELIVERY ANALYSIS │
│ recording_2025_01_15.mp4 │
├────────────────────────────────────────┤
│ OVERALL SCORE: 18/25 (Competent) │
│ │
│ Content & Organization: 4/5 │
│ Delivery & Vocal Quality: 3/5 │
│ Body Language & Eye Contact: 4/5 │
│ Audience Engagement: 4/5 │
│ Language & Clarity: 3/5 │
└────────────────────────────────────────┘
# Video Delivery Analysis
**File:** recording_2025_01_15.mp4
**Duration:** 12:34
**Date:** 2025-01-15
---
## VOICE ANALYSIS
| Metric | Value | Target | Assessment |
|--------|-------|--------|------------|
| Speech Rate | 145 WPM | 120-160 | ✅ Good |
| Pitch Variation | 42.3 Hz | >30 Hz | ✅ Engaging |
| Volume Consistency | 0.08 | <0.15 | ✅ Steady |
**Filler Words Detected:**
- "um" - 8 times
- "you know" - 5 times
- "basically" - 3 times
**Recommendation:** Reduce "um" usage. Try pausing instead.
---
## FACIAL ANALYSIS
| Metric | Value | Assessment |
|--------|-------|------------|
| Eye Contact Frequency | 72% | ✅ Good |
| Smile Frequency | 35% | ⚠️ Could increase |
**Emotion Timeline:**
- 0:00-2:00: Neutral (intro)
- 2:00-8:00: Happy/Engaged (main content)
- 8:00-10:00: Serious (data presentation)
- 10:00-12:34: Happy (conclusion)
**Recommendation:** More smiles during technical sections.
---
## CONTENT ANALYSIS
**Strengths:**
- Clear opening hook
- Good use of clinical examples
- Strong call-to-action
**Areas for Improvement:**
- Could use more pauses after key points
- Consider adding more Hinglish transitions
- Section on side effects could be more structured
---
## OVERALL FEEDBACK
**What You Did Well:**
1. Excellent pace - not too fast, not too slow
2. Good eye contact with camera
3. Clinical examples were relatable
**What to Improve:**
1. Reduce filler words (especially "um")
2. Add more smiles during technical explanations
3. Pause after key statistics for emphasis
**Score: 18/25 - Competent Speaker**
You're delivering solid content with room for refinement.
This skill is calibrated for Hinglish content:
| Feature | What It Checks |
|---|---|
| Code-switching | Natural Hindi ↔ English transitions |
| Pace adjustment | Slower for English technical terms |
| Cultural markers | Use of "ji", "beta", "aapko bata doon" |
| Engagement phrases | "Dekho", "Suniye", "Samjhe?" |
Track your improvement across recordings:
┌─────────────────────────────────────────────────────┐
│ PROGRESS TRACKER (Last 5 Videos) │
├─────────────────────────────────────────────────────┤
│ Video │ Score │ Main Improvement │
│ ───────────────────────────────────────────────── │
│ Jan 10 │ 15/25 │ Baseline │
│ Jan 15 │ 18/25 │ Better eye contact │
│ Jan 20 │ 17/25 │ Fewer filler words │
│ Jan 25 │ 19/25 │ More varied pace │
│ Jan 30 │ 21/25 │ Natural Hinglish flow │
└─────────────────────────────────────────────────────┘
Record Video → Analyze with video-delivery-coach → Fix issues → Re-record (optional) → Publish
youtube-script-master - Script adjustments based on delivery feedback# Core (required)
pip install anthropic python-dotenv rich
# Voice analysis
pip install librosa moviepy faster-whisper
# Facial analysis (optional - for full analysis)
pip install opencv-python mediapipe deepface tf-keras
# Note: tf-keras is heavy (~500MB). Skip for voice-only mode.
| Key | Purpose | Status |
|---|---|---|
| ANTHROPIC_API_KEY | Final analysis and coaching | Already have |
python scripts/analyze_video.py --video file.mp4 --voice-only
python scripts/analyze_video.py --video file.mp4 --full
When user asks to analyze a video:
import os
if not os.path.exists(video_path):
print("Video file not found")
return
python scripts/analyze_video.py --video "/path/to/video.mp4"
=== VIDEO DELIVERY ANALYSIS ===
File: hinglish_statin_video.mp4
Duration: 15:23
VOICE METRICS:
├── Speech Rate: 138 WPM (Target: 120-160) ✅
├── Pitch Variation: 38.5 Hz ✅ Natural variation
└── Volume: Consistent ✅
FILLER WORDS:
├── "um": 12 occurrences
├── "basically": 8 occurrences
└── "you know": 5 occurrences
FACIAL METRICS:
├── Eye Contact: 68% ✅ Good
├── Smiles: 28% ⚠️ Below target (40%)
└── Dominant Emotion: Engaged
CONTENT SCORE:
├── Content & Organization: 4/5
├── Delivery & Vocal Quality: 3/5
├── Body Language: 4/5
├── Engagement: 4/5
└── Language & Clarity: 4/5
TOTAL: 19/25 (Proficient Speaker)
TOP 3 IMPROVEMENTS:
1. Replace "um" with pauses
2. Smile more during technical explanations
3. Slow down slightly when explaining statistics
HINGLISH NOTES:
✅ Natural code-switching
✅ Good use of "aapko batata hoon"
⚠️ Consider more "samjhe?" checks for engagement
This skill helps you improve your delivery over time - not by judging, but by giving you objective data to work with.