一键在 Manus 中运行任何 Skill

youtube-researcher

星标4

分支1

更新时间2026年5月12日 18:09

Use when the user wants YouTube research at speed: finding relevant videos in a niche, scanning channels, pulling transcripts, comparing competitors, or gathering source material from YouTube before summarizing trends, hooks, topics, or positioning. This skill uses a local helper script wired to SerpApi for YouTube search and Supadata for fast transcripts and channel data.

安装

用 Codex 或 Claude 帮你安装复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。

在 Manus 中运行

来源

rbrown101010

rbrown101010/Rileys-Top-Skills

打开 GitHub 仓库查看创作者相关仓库

下载

在 Manus 中运行

YouTube Researcher

Overview

Use this skill for YouTube topic research, niche scouting, competitor scans, channel review, and transcript collection. Prefer the bundled script over ad hoc API calls so the workflow stays fast, repeatable, and consistent across chats.

Quick Start

Run the helper script directly:

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py --help

Common commands:

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py search "ai coding agents" --limit 10 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py search "b2b saas onboarding" --limit 8 --transcripts 5 --excerpt-chars 900 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py channel "@ycombinator" --limit 12 --with-transcripts 3 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py transcript "https://www.youtube.com/watch?v=dQw4w9WgXcQ" --lang en

Workflow

1. Search the niche

Start with search when the user is exploring a topic, niche, idea cluster, or competitor landscape.

Use --limit to pull more videos.
Use --channel "name" to keep only results whose channel name matches a substring.
Use --sp when the user wants a YouTube-native filter token copied from a filtered YouTube URL.
Use --no-cache for fast-changing topics when freshness matters.

2. Pull transcripts only where useful

Transcripts cost more requests than search, so avoid fetching them for everything unless the user clearly needs them.

Use --transcripts N on search to fetch transcripts for the top N search results.
Use transcript for one-off deep dives.
Use --mode native if the user wants only existing captions and wants to avoid AI-generated fallback.
Use --mode auto for the default fast path.

3. Inspect channels directly

Use channel when the user wants to understand a creator, competitor, or niche-specific publisher.

It fetches Supadata channel metadata plus recent video IDs.
It enriches returned videos with metadata so titles, views, dates, and thumbnails are available without extra work.
Use --with-transcripts N to attach transcript excerpts for the first N channel videos.

4. Synthesize in-chat

After using the script, do the analysis in the chat:

recurring hooks or thumbnail/title patterns
topic clusters and content gaps
channel positioning
common claims, framing, and audience promises
relevant videos worth deeper review

Output Guidance

Prefer --json when the result will be analyzed further in the chat.
Prefer plain output when the user mainly wants a quick list or transcript excerpt.
For “latest” or time-sensitive requests, pair --no-cache with concrete dates in the final answer.

Keys

This public version does not include API keys. Set the required credentials in the shell environment before running the helper:

SUPADATA_API_KEY
SERPAPI_KEY
SERPAPI_API_KEY or SERP_API_KEY also work for SerpApi

Do not write API keys into this skill file or the helper script.

References

references/api-notes.md for the current endpoint notes and parameter reminders

同仓库更多 Skills

同仓库

brand-deal-researcher

rbrown101010/Rileys-Top-Skills

Find paid promotion, sponsorship, brand deal, influencer campaign, creator partnership, affiliate, and paid collaboration opportunities in the user's Gmail or email exports; research each brand/client; dedupe follow-ups; prioritize fit; and produce a Markdown table, document, or spreadsheet of brand leads with notes. Use when the user asks to scan recent email for brand deals, paid promo offers, sponsorship pitches, creator partnerships, or to turn those leads into a table/spreadsheet/doc.

2026-05-124

buffer-publisher

rbrown101010/Rileys-Top-Skills

Use Buffer's GraphQL API to inspect Buffer accounts, organizations, channels, queues, posts, daily limits, and to create Buffer Ideas, drafts, scheduled posts, queued posts, or media posts. Use when the user asks to draft, schedule, queue, publish, edit, delete, review, list, or manage social posts through Buffer, or asks what is connected in their Buffer account.

2026-05-124

chorus-ios

rbrown101010/Rileys-Top-Skills

Build, preview, sign, install, and publish iOS apps from Codex using the Vibecode/Chorus signing CLI. Use when the user wants to build an iOS app from source, create a simulator preview, sign an iOS app, install on their iPhone, register a device, set up Apple Developer auth, publish to the App Store or TestFlight, submit for review, or work with the iOS build/sign/distribute/publish pipeline. Also triggers for "test on device", "put this on my phone", "ship it to my iPhone", "publish to App Store", "submit my app", "push to TestFlight", "send for review", or any request to get a Swift/iOS app running on a real device or shipped to App Store Connect.

2026-05-124

excalidraw-diagrams

rbrown101010/Rileys-Top-Skills

Create polished Excalidraw diagrams from text descriptions. Use when the user asks for a diagram, flowchart, architecture diagram, decision tree, process flow, data flow, system overview, or similar visual explanation, or says "draw this", "diagram this", "visualize this", or "make a chart". Produces styled `.excalidraw` files with consistent colors and can export a shareable excalidraw.com URL.

2026-05-124

genmedia

rbrown101010/Rileys-Top-Skills

Use the genmedia CLI to search, inspect, run, and manage 1200+ fal.ai model endpoints. Trigger when the user mentions "genmedia", "fal CLI", or asks to "search models", "run a model", "fetch schema", "check pricing", "upload to fal", "queue async job", "track request", or any direct interaction with the fal.ai endpoint catalog. This is the foundational skill. Every other fal.ai-related skill in this repo executes its work through genmedia commands. Use `--json` whenever the output will be parsed by an agent.

2026-05-124

internet-image-puller

rbrown101010/Rileys-Top-Skills

Use when Codex needs to find images on the public internet and save them into a local directory, especially for brand assets, moodboards, research boards, reference pulls, or site-specific image collection. Supports SerpAPI Google Images search for query-driven discovery and Firecrawl scrape/map workflows for extracting images from specific pages or whole sites.

2026-05-124

name	youtube-researcher
description	Use when the user wants YouTube research at speed: finding relevant videos in a niche, scanning channels, pulling transcripts, comparing competitors, or gathering source material from YouTube before summarizing trends, hooks, topics, or positioning. This skill uses a local helper script wired to SerpApi for YouTube search and Supadata for fast transcripts and channel data.

YouTube Researcher

Overview

Quick Start

Run the helper script directly:

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py --help

Common commands:

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py search "ai coding agents" --limit 10 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py search "b2b saas onboarding" --limit 8 --transcripts 5 --excerpt-chars 900 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py channel "@ycombinator" --limit 12 --with-transcripts 3 --json

python3 $HOME/.codex/skills/youtube-researcher/scripts/youtube_research.py transcript "https://www.youtube.com/watch?v=dQw4w9WgXcQ" --lang en

Workflow

1. Search the niche

Start with search when the user is exploring a topic, niche, idea cluster, or competitor landscape.

Use --limit to pull more videos.
Use --channel "name" to keep only results whose channel name matches a substring.
Use --sp when the user wants a YouTube-native filter token copied from a filtered YouTube URL.
Use --no-cache for fast-changing topics when freshness matters.

2. Pull transcripts only where useful

Transcripts cost more requests than search, so avoid fetching them for everything unless the user clearly needs them.

Use --transcripts N on search to fetch transcripts for the top N search results.
Use transcript for one-off deep dives.
Use --mode native if the user wants only existing captions and wants to avoid AI-generated fallback.
Use --mode auto for the default fast path.

3. Inspect channels directly

Use channel when the user wants to understand a creator, competitor, or niche-specific publisher.

It fetches Supadata channel metadata plus recent video IDs.
It enriches returned videos with metadata so titles, views, dates, and thumbnails are available without extra work.
Use --with-transcripts N to attach transcript excerpts for the first N channel videos.

4. Synthesize in-chat

After using the script, do the analysis in the chat:

recurring hooks or thumbnail/title patterns
topic clusters and content gaps
channel positioning
common claims, framing, and audience promises
relevant videos worth deeper review

Output Guidance

Prefer --json when the result will be analyzed further in the chat.
Prefer plain output when the user mainly wants a quick list or transcript excerpt.
For “latest” or time-sensitive requests, pair --no-cache with concrete dates in the final answer.

Keys

This public version does not include API keys. Set the required credentials in the shell environment before running the helper:

SUPADATA_API_KEY
SERPAPI_KEY
SERPAPI_API_KEY or SERP_API_KEY also work for SerpApi

Do not write API keys into this skill file or the helper script.

References

references/api-notes.md for the current endpoint notes and parameter reminders