keyapi-youtube-video-analysis（keyapi youtube 视频分析）

keyapi-youtube-video-analysis

Analyze YouTube videos at depth — from full metadata and stream formats to comment threads, related recommendations, Shorts discovery, and trending video intelligence.

This skill provides comprehensive YouTube video intelligence using the KeyAPI MCP service. It enables detailed video inspection (metadata, views, engagement), layered comment analysis (top-level comments + nested replies), playback stream retrieval, related content discovery, Shorts-specific search, keyword-based video search, and trending video monitoring — all through a cache-first workflow.

Use this skill when you need to:

• Retrieve complete metadata, engagement metrics, and raw player data for any YouTube video
• Read and analyze top-level comments and their nested reply threads
• Obtain all available playback stream URLs and quality levels for a video
• Discover related and recommended videos as surfaced by the YouTube algorithm
• Search YouTube Shorts by keyword with category and time-based filtering
• Find videos by keyword with advanced sort and recency filters
• Monitor trending videos globally or by category (Music, Gaming, Movies)

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Prerequisites

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

MCP Server Configuration

All tool calls in this skill target the KeyAPI YouTube MCP server:

Server URL : https://mcp.keyapi.ai/youtube/mcp
Auth Header: Authorization: Bearer $KEYAPI_TOKEN

Setup (one-time):

# 1. Install dependencies
npm install

# 2. Set your API token (get one free at https://keyapi.ai/)
export KEYAPI_TOKEN=your_token_here

# 3. List all available tools to verify the connection
node scripts/run.js --platform youtube --list-tools

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Analysis Scenarios

Video Data Nodes

Search & Discovery Nodes

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Workflow

Step 1 — Identify the Video Analysis Objective and Select Nodes

Clarify the research goal and map it to one or more nodes. Common patterns:

• Video deep-dive: Start with get_video_information (set need_format: true for clean structured output), then layer get_video_comments and get_video_streams_info.
• Comment thread analysis: get_video_comments → collect reply_continuation_token from comments with replies → get_video_sub_comments per thread.
• Related content mapping: get_related_videos returns the full recommendation set in one call — no pagination needed.
• Shorts research: youtube_shorts_search with keyword → use continuation_token for pure Shorts on subsequent pages.
• Trend monitoring: get_trending_videos with section filter → cross-reference with search_video by keyword.

Video ID extraction Extract the video ID from the URL: - https://www.youtube.com/watch?v=oaSNBz4qMQY → oaSNBz4qMQY Most endpoints accept either video_id (recommended) or video_url (full URL string) — use video_id where possible for reliability.

Step 2 — Retrieve API Schema

Before calling any node, inspect its input schema to confirm required parameters and available options:

node scripts/run.js --platform youtube --schema <tool_name>

# Examples
node scripts/run.js --platform youtube --schema get_video_information
node scripts/run.js --platform youtube --schema youtube_shorts_search

Step 3 — Call APIs and Cache Results Locally

Execute tool calls and persist responses to the local cache.

Calling a tool:

node scripts/run.js --platform youtube --tool <tool_name> \
  --params '<json_args>' --pretty

# Skip cache for fresh results
node scripts/run.js --platform youtube --tool <tool_name> \
  --params '<json_args>' --no-cache --pretty

Example — get full video information (formatted):

node scripts/run.js --platform youtube --tool get_video_information \
  --params '{"video_id":"oaSNBz4qMQY","need_format":true,"language_code":"en-US"}' --pretty

Example — get top comments:

node scripts/run.js --platform youtube --tool get_video_comments \
  --params '{"video_id":"oaSNBz4qMQY","sort_by":"top","language_code":"en-US","country_code":"US"}' --pretty

Example — paginate to next comment page:

node scripts/run.js --platform youtube --tool get_video_comments \
  --params '{"video_id":"oaSNBz4qMQY","continuation_token":"<token_from_previous_response>"}' --pretty

Example — get replies to a comment:

# Extract reply_continuation_token from a comment in get_video_comments response
node scripts/run.js --platform youtube --tool get_video_sub_comments \
  --params '{"continuation_token":"<reply_continuation_token>","language_code":"en-US"}' --pretty

Example — get related videos:

node scripts/run.js --platform youtube --tool get_related_videos \
  --params '{"video_id":"dQw4w9WgXcQ","need_format":true}' --pretty

Example — search Shorts:

# First call (may return mixed content)
node scripts/run.js --platform youtube --tool youtube_shorts_search \
  --params '{"search_query":"cooking tips","filter_mixed_content":true,"language_code":"en-US","country_code":"US"}' --pretty

# Subsequent pages (pure Shorts via continuation_token)
node scripts/run.js --platform youtube --tool youtube_shorts_search \
  --params '{"search_query":"cooking tips","continuation_token":"<token>"}' --pretty

Example — search videos with recency filter:

node scripts/run.js --platform youtube --tool search_video \
  --params '{"search_query":"AI tutorial","order_by":"this_month","country_code":"US","language_code":"en"}' --pretty

Example — get trending videos:

node scripts/run.js --platform youtube --tool get_trending_videos \
  --params '{"section":"Music","country_code":"US","language_code":"en"}' --pretty

Pagination:

get_video_sub_comments critical note: This endpoint requires the reply_continuation_token field extracted from a specific comment in the get_video_comments response — it is different from the comment-level continuation_token used to paginate comments. Do NOT pass the video_id to this endpoint.

get_video_streams_info time limit: Returned playback URLs are valid for approximately 6 hours. Use them promptly. HD video (720p+) streams require separate audio and video downloads that must be merged locally.

youtube_shorts_search mixed content: The first request may return a mix of Shorts and regular videos. Use continuation_token on subsequent calls for pure Shorts results. Set filter_mixed_content: true (default) to filter long videos from the first response.

Cache directory structure:

.keyapi-cache/
└── YYYY-MM-DD/
    ├── get_video_information/
    │   └── {params_hash}.json
    ├── get_video_comments/
    │   └── {params_hash}.json
    ├── get_video_sub_comments/
    │   └── {params_hash}.json
    ├── get_video_streams_info/
    │   └── {params_hash}.json
    ├── get_related_videos/
    │   └── {params_hash}.json
    ├── youtube_shorts_search/
    │   └── {params_hash}.json
    ├── search_video/
    │   └── {params_hash}.json
    └── get_trending_videos/
        └── {params_hash}.json

Cache-first policy:

Before every API call, check whether a cached result already exists. If valid, load from disk and skip the API call.

Step 4 — Synthesize and Report Findings

For video analysis:

1. Video Overview — Title, channel, publish date, view count, like count, comment count, duration, tags, description summary.
2. Engagement Quality — Like-to-view ratio, comment depth (replies per top-level comment), engagement rate estimate.
3. Comment Intelligence — Top themes from most-liked comments, sentiment signals, key community questions or reactions.
4. Related Content Map — Algorithm-recommended videos, common topics and channels in the recommendation cluster.
5. Stream Quality Summary — Available resolutions, format types (merged vs. adaptive), playback URL validity window.

For search / discovery:

1. Search Results Overview — Video count returned, top titles, channel distribution, view spread.
2. Shorts Landscape — Trending Shorts topics, creator diversity, engagement patterns.
3. Trending Signals — Top trending videos by category, creator names, view velocity indicators.

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Common Rules

author: KeyAPI license: MIT repository: https://github.com/EchoSell/keyapi-skills

Error Handling