YouTube Research

Follow shared release-shell rules in:

• postplus-shared release-shell rules

Use this skill when the request is about YouTube channels, audience proxies, or public videos.

This skill is for:

• collecting channel metadata and subscriber counts
• collecting audience comment samples from a specific video
• collecting one or more public YouTube video URLs
• discovering a small YouTube video set from a broad query
• normalizing YouTube video metrics into local artifacts
• writing local raw payloads, normalized datasets, and short markdown summaries

This skill is not for:

• LinkedIn collection
• Facebook collection
• TikTok, Instagram, or X workflows
• publishing, deleting, or scheduling posts
• promising a subscriber list that YouTube does not expose publicly

Hosted Capability Boundary

This skill depends on host-managed collection capability for the corresponding collection keys.

In the product shell:

• do not probe or print provider secrets
• do not ask the user to export them inside chat
• if a collection key returns a stable capability/network hard error, stop immediately instead of trying alternate shell commands

Default Collection Keys

Use these hosted collection keys by default:

• channel metadata and subscriber counts

  • collection key: youtube-channel-summary
  • use for channel metadata and subscriber counts
  • tested input shape:
    • channels = ["@Google"]
    • maxVideosPerChannel = 0
    • includeChannelInfo = true
    • includeVideos = false
  • current observed output fields include:
    • channelId
    • channelName
    • channelHandle
    • subscriberCount
    • totalVideos
    • avatarUrl
    • bannerUrl

• audience comments collection

  • collection key: youtube-comments
  • use for audience proxy research from comments on a specific video or Shorts URL
  • tested input shape:
    • startUrls = ["https://www.youtube.com/watch?v=<video-id>"]
    • sort
    • maxItems
  • current observed output fields include:
    • id
    • text
    • likeCount
    • replyCount
    • publishedTime
    • author

• downloadable video records

  • collection key: youtube-video-download
  • use when the workflow needs the rented downloader-backed record for explicit video URLs

Use comments as an audience proxy when subscriber identities are not public.

Failure Posture

• fail if the request includes non-YouTube platforms
• fail if no YouTube public URLs can be discovered
• fail if hosted content collection returns malformed items without stable URL or id fields
• fail if the channel collection cannot resolve the channel handle or URL
• keep raw responses for debugging

Recommended Workflow

1. For channel research, start with youtube-channel-summary.
2. For audience research, collect a small comments sample from representative videos with youtube-comments.
3. For broad public video discovery, use the hosted content collection scripts below.

Do not present comment authors as a full subscriber list.

Release-Shell Execution Contract

• keep collection briefs, raw datasets, normalized outputs, and summary caches under <work-folder>/.postplus/youtube-research/
• keep only final user-facing summaries or shortlist exports outside .postplus/
• compile a small channel, comment, or public-video brief before the expensive collection step
• start with a bounded first pass:
  • one channel batch
  • one comments batch
  • one public video plan

Public Video Implementation

Read before implementation:

• skills/20-research/youtube-research/references/normalized-schema.md

Use these entrypoints:

• skills/20-research/youtube-research/scripts/run_youtube_video_collection.mjs

Hosted Collection Note

Use the shared hosted collection runner for actor calls:

• ${CLAUDE_SKILL_DIR}/_postplus_shared/00-core/shared-collection/scripts/collection_actor_run.mjs