yt-research
Skill4 starsChanged 4 months agoMITView on GitHub
Research YouTube for content strategy — find what's taking off in a niche, scan competitor channels, rank videos by views and by view-velocity, read comments, see trending Shorts, and pull any video's transcript. Use this WHENEVER you need to research YouTube, find trending/viral videos on a topic, see what's blowing up on YT, scan a competitor's channel, get a YouTube video's transcript or 'what's in this video', or gather YouTube content ideas from what's working — even if the skill isn't named. Data comes from Loore (LOORE_API_KEY); the official YouTube Data API v3 (YOUTUBE_API_KEY) is an optional alternative for search, channels and comments. Never scrape YouTube from this machine.
- Reads credentials
2 other projects use a version of this.
Adopted by
Most recently taken up by
- mategog/mateai-youtube-mcptheir own version, first seen Nov 15, 2026
- utillayer/youtube-agent-skilltheir own version, first seen Nov 8, 2026
Other versions2 projects use a version of this.
Word for word · 0
None found.
Edited · 0
None found.
Wrote their own version · 2
- utillayer/youtube-agent-skill★ 10same name
- mategog/mateai-youtube-mcp★ 1same name
How matches are found. Word for word: the exact same text. Edited: shares passages with this file (“99% same” means 99% of its five-word phrases also appear here; 10% or more counts). Their own version: similar meaning, or, weakest, the same name. We can’t tell which project had it first. Projects are counted among the 200 most-starred candidates, and mentions are links to the repository we found, up to 50 per site, so both counts are a floor.
In this repository43 other files it gives its agents.
- agent-browserSkill
- agent-introspectionSkill
- agentos-contentSkill
- carousel-instagramSkill
- chat-archiveSkill
- codex-imageSkill
- content-engineSkill
- cross-reviewSkill
- datawrapperSkill
- dev-pipelineSkill
- excalidrawSkill
- fast-loop-codingSkill
- groq-voiceSkill
- higgsfield-generateSkill
- higgsfield-marketplace-cardsSkill
- higgsfield-product-photoshootSkill
- higgsfield-soul-idSkill
- instagram-superpowerSkill
- learningsSkill
- loop-codingSkill
- markdown-newSkill
- mcp-api-buildSkill
- mcp-builderSkill
- memory-auditSkill
- miro-boardSkill
- perplexity-researchSkill
- presentSkill
- reel-radarSkill
- reels-analytics-for-brokersSkill
- reelsSkill
- remote-agent-desktopSkill
- security-auditSkill
- senior-brainstormSkill
- seo-tuneSkill
- server-doctorSkill
- telegram-chipSkill
- threads-contentSkill
- topic-monitorSkill
- transcriptSkill
- twitterSkill
- workshopSkill
- youtube-producerSkill
- youtube-thumbnailSkill
The file
We split this skill into the rules it gives an agent
7 rules. About 0 are common, 4 are distinctive, and 3 are very specific to this project.
≈0
Common · estimate
Many files like this have a rule of this kind.
≈4
Distinctive · estimate
This author's own choices. Worth a look.
3
Very specific
Tied to their project's names, files or tools. Probably skip.
- Common
- Distinctive
- Very specific
- Distinctive: line 17
Do not scrape YouTube from your own IP — no yt-dlp, innertube, oEmbed or transcript sites for data.
- Distinctive: line 26
Spend approval: Every Loore YouTube call is paid (search, channel, video, comments, trending: 1.5 credits per call.
- Distinctive: line 66
Search dates are approximate (YouTube shows «3 weeks ago»).
- Very specific: line 96
Keep a persistent base of tracked channels in references/watchlist.json (ships as a template — fill it with your own niche).
- Very specific: line 105
Add a channel: read it with Loore yt_channel (by handle, UC… id or any channel link) — name, subscribers, video totals, country, keywords:
- Distinctive: line 136
No captions → exit 4 and the list of caption languages the video has.
- Very specific: line 161
Default — a numbered list: views | velocity/day | duration — title + a channel — url line.
≈ “Common” is an estimate: at least 1 in 5 of the 364,980 skill files we read have a rule of the same kind, not the same words. “Very specific” is our guess from the words: it names a file, a command, a setting or the project itself.
Files
yt-research — YouTube niche & competitor radar
Find what's taking off on YouTube in your niche, scan competitor channels, rank by virality,
read comments and pull transcripts for analysis. Pairs well with an Instagram radar (e.g.
reel-radar) and an X radar — this is the YouTube leg.
Where the data comes from. YouTube data is read only through Loore
(default, docs: https://docs.loore.ai) or the official YouTube Data API v3 (optional --api).
Do not scrape YouTube from your own IP — no yt-dlp, innertube, oEmbed or transcript sites for
data. yt-dlp is acceptable only to download a video file itself, and this skill does not need that.
Setup
export LOORE_API_KEY="..." # https://app.loore.ai → Настройки → API и MCP (shown once; never in chat or git)Spend approval. Every Loore YouTube call is paid (search, channel, video, comments, trending: 1.5 credits per call; transcript: 2 credits, only when one is returned). An empty page or an upstream error is free. State the total cap in credits to the user once, and after their «yes» create a flow budget (valid 24 hours):
python3 scripts/_loore.py approve --flow yt-research --cap 100
export LOORE_SPEND_AUTHORIZATION_ID=<printed id>The scripts send spend_authorization_id with every paid call, including every next page. Without
it Loore answers 428 needs_user_confirmation. On 429 the client waits Retry-After; when the
window resets later than 120 seconds (daily budget), it stops with exit code 4 and prints when to rerun.
MCP instead of scripts. The same operations are MCP tools on https://mcp.loore.ai/mcp
(header Authorization: Bearer <key>): yt_search, yt_channel, yt_channel_videos,
yt_channel_shorts, yt_video, yt_comments, yt_transcript, yt_trending_shorts
(yt_search ↔ POST /v1/yt-search). Full list: https://docs.loore.ai/mcp/tools.
When to use
- "What's taking off on YouTube about
<niche>?" - "Scan these competitor channels — what's working for them?"
- "What are people saying under this video?"
- "Get the transcript / what's in this video
<url>?" - Gathering proven content ideas before planning a video.
1. Niche search + ranking (Loore yt_search)
python3 scripts/yt_search.py "<query>" --limit 15 # top by views
python3 scripts/yt_search.py "AI agents for business" --by velocity --enrich # what's accelerating
python3 scripts/yt_search.py "Claude Code" --limit 20 --json # raw JSON for downstream
python3 scripts/yt_search.py "Claude Code" --upload-date this_week --sort popular --region US
python3 scripts/yt_search.py "AI agents" --type shorts --limit 20--by views(default) — top by absolute views.--by velocity— by view-velocity (views / days since upload). The best "rising right now" signal: a fresh video with high daily velocity matters more than an old one with big-but-slow totals. Search dates are approximate (YouTube shows «3 weeks ago»); rows without a date are enriched automatically.--enrich— exact upload date, likes, comments (yt_video) and channel subscribers (yt_channel, once per channel). One paid call per video — run on short lists.- Filters map to
yt_searchparameters:--upload-date(today / this_week / this_month / this_year),--sort(relevance / popular),--length(under_3_min / between_3_and_20_min / over_20_min),--region(two-letter code),--type(videos / shorts). - Pages follow
nextCursoruntil--limitis reached (at most 5 pages per search or channel).
2. Scan competitor channels (Loore yt_channel_videos / yt_channel_shorts)
python3 scripts/yt_search.py --handles "channelA,channelB" --limit 10 --by velocity --enrich
python3 scripts/yt_search.py --handles "channelA" --shorts --limit 20 # the channel's ShortsPulls the latest N long-form videos (or, with --shorts, the newest Shorts) of each channel and
ranks them — so you see which formats are firing for a competitor right now. Handle with or without
@, or a UC… channel id.
Trending Shorts (Loore yt_trending_shorts)
python3 scripts/yt_search.py --trending-shorts --limit 48About 48 Shorts trending on YouTube right now, one call.
Channel base — references/watchlist.json
Keep a persistent base of tracked channels in references/watchlist.json (ships as a template —
fill it with your own niche). Each entry: handle, name, url, channel_id, subs, lang, niche, cadence,
note, added. Run the radar over the whole base in one line:
HANDLES=$(python3 -c "import json; d=json.load(open('references/watchlist.json')); print(','.join(c['handle'] for c in d['channels']))")
python3 scripts/yt_search.py --handles "$HANDLES" --by velocity --enrich --limit 8Add a channel: read it with Loore yt_channel (by handle, UC… id or any channel link) —
name, subscribers, video totals, country, keywords:
python3 scripts/_loore.py call yt-channel '{"url": "https://www.youtube.com/@name"}'then append an object to channels[] (niche from recent titles, cadence from recent upload dates).
3. Comments under a video (Loore yt_comments)
python3 scripts/yt_search.py --comments "<youtube_url|id>" --order top --limit 30
python3 scripts/yt_search.py --comments "<youtube_url|id>" --order newest --jsonOutput: likes | replies — author: text. With --json each comment also carries replies_token;
read a thread with python3 scripts/_loore.py call yt-comments '{"replies_token": "<token>"}'.
4. Transcript of a video (Loore yt_transcript)
bash scripts/yt_transcript.sh "<youtube_url|id>" [text|json]- Public captions of any video or Short, any length (no AI transcription). Long transcripts come
in windows (
truncated/nextOffset); the script reads them all. textprints a# title — channelheader (oneyt_videocall) and the plain transcript;jsonprints{video_id, language, transcript, metadata}.- More options (timestamps, caption language, original audio only, no header) —
python3 scripts/yt_transcript.py --help. - No captions → exit 4 and the list of caption languages the video has.
- The official YouTube Data API cannot download captions of other people's videos (
captions.downloadworks only for videos you can edit), so there is no API alternative for transcripts.
Then extract key points, compare competitors' narratives, and build the structure for your own video.
5. YouTube Data API v3 (optional alternative)
Official path for search, channel videos and comments. Enable with --api; reads YOUTUBE_API_KEY
from the environment and sends it in the X-goog-api-key header (never in a URL or command line).
export YOUTUBE_API_KEY=...
python3 scripts/yt_search.py "AI agents" --api --limit 20 # search.list + videos.list
python3 scripts/yt_search.py --handles "channelA" --api --limit 20 # channels.list + playlistItems.list + videos.list
python3 scripts/yt_search.py --comments "<url>" --api --order newest # commentThreads.listSetup (once): in Google Cloud, enable YouTube Data API v3 → create an API key → export
YOUTUBE_API_KEY. Free quota: 10,000 units/day (search.list = 100 units/request ≈ 100 searches/day;
channel, video and comment reads ≈ 1 unit each). Not available on this path: Shorts lists, trending
Shorts and transcripts (use Loore).
Output
Default — a numbered list: views | velocity/day | duration — title + a channel — url line.
With --json — an array of objects (id, title, channel, views, likes, comments, duration,
upload_date, velocity, url, plus channel_subs after --enrich) for downstream analysis or
writing to your base.
How to read it (content strategy)
- High velocity on a fresh video = a topic on the rise → candidate for your plan now.
- High absolute views on an old one = evergreen → a base topic, not urgent.
- Compare the winners' format / title / length — that's what works in the niche.
- Transcript of a top video → extract the structure and points; don't copy, do it better in your voice.
Limits & notes
- Public data only — no private analytics of someone else's channel (only the owner has that).
--enrich/--by velocitymake one paid call per video — run on short lists, in batches.- Private analytics of YOUR OWN channel needs OAuth (YouTube Analytics API) — separate setup.
- Loore errors:
401key,402out of credits,403key not allowed this method,428spend approval needed — the script stops with a hint;429waits or stops with a rerun time. - The channel base (
references/watchlist.json) ships as a template — populate it with your niche.
Does it work?
Reviews from people and from AI agents that ran it: what they used it for, and whether it worked.
No reviews yet
People and AI agents can review whether this works in practice. Say what you used it for, and whether it worked.
- It worked
- Partly
- It didn't
Using an agent? It can post a review for you after it has run the file: the MCP tool registry_write, action report. How to connect one.
Your trail
Files you open will collect here, so you can find them again.
Open two or more, and you can remix the rules you like from each into one list. Kept in this browser only.
Now do this for your own.
Bring your CLAUDE.md, AGENTS.md or skills. agentleFS shows which of your rules are common, which are yours and which only fit one project, and keeps one version your agents read everywhere.

