Skill: trend-scout-global

Domain: Global short-video and social trend monitoring (TikTok / Instagram Reels / YouTube Shorts / X / Pinterest / CapCut Templates) Version: 1.0.0 Date: 2026-04-13

Objective

Discover, validate, clean, structure, grade, and report global social-media trend signals across TikTok, Instagram Reels, YouTube Shorts, X, Pinterest, and CapCut Templates.

Output Schema

Every trend record MUST conform to the following schema:

{
  "trend_id": "string",
  "platform": "TikTok|Instagram|YouTube|X|Pinterest|UNKNOWN",
  "region": "US|JP|BR|UK|DE|IN|PH|AU|UNKNOWN",
  "content_type": "hashtag|audio|video|effect",
  "name": "#canonical hashtag",
  "rank": 1,
  "views_or_posts": 0,
  "growth_7d": 0.0,
  "related_hashtags": ["string"],
  "source_url": "https://...",
  "crawl_time": "2026-04-13T02:40:00Z"
}

All timestamps must be ISO-8601 UTC ending with Z.

Cross-Platform Information Source Understanding

• Platform Detection: Identify the originating platform from source_url, explicit platform fields, or language hints.
  • tiktok.com / ads.tiktok.com / 日本版TikTok → TikTok
  • instagram.com / reels → Instagram
  • youtube.com / youtube_shorts / YouTube Shorts → YouTube
  • x.com / twitter.com → X
  • pinterest.com → Pinterest
  • capcuttemplate.io / capcutonline.com → CapCut
  • Unrecognized aggregator → UNKNOWN
• Region Mapping: Map region hints to ISO-style codes.
  • Japan / 日本 / ja-JP → JP
  • Brasil / Brazil → BR
  • UK / United Kingdom → UK
  • DE / Germany → DE
  • IN / India → IN
  • PH / Philippines → PH
  • AU / Australia → AU
  • US / United States → US
  • Missing or ambiguous → UNKNOWN
• Content Type: Derive from context.
  • Hashtag entries → hashtag
  • Video/Shorts entries → video
  • Pinterest boards / CapCut templates / effects → effect
  • Audio/BGM trends → audio

Trend Recognition & Theme Extraction

• Canonical Name Generation: Convert raw titles to canonical English hashtags.
  • Spanish Amor de Vampiro → #vampire love
  • Second Chance Romance → #second chance
  • Mafia Romance Aesthetic → #mafia romance
  • Werewolf Mate → #werewolf mate
  • Contract Marriage → #contract marriage
  • Strong Female Lead → #strong female lead
  • Vertical Drama → #vertical drama
  • Interactive Story → #interactive story
• IP-Style Themes: Detect associated themes: billionaire romance, werewolf mate, mafia romance, strong female lead, revenge, vampire love, contract marriage, second chance, interactive story, vertical drama.
• Related Hashtags: Extract without duplication. Output [] (empty array) when none exist. Never output null for related_hashtags.

Data Cleaning & Structuring

• Numeric Parsing:
  • Convert strings like "1.3B" → 1300000000, "24.5B" → 24563880969, "45.4B" → 45416623921.
  • Convert percentage strings "+24%" → 0.24, "+298%" → 2.98, "+106%" → 1.06 for growth_7d.
  • If growth is missing, default growth_7d to 0 or 1 depending on context (stable trend default).
• Required Fields: Ensure trend_id, platform, region, content_type, name, rank, views_or_posts, growth_7d, related_hashtags, source_url, crawl_time are always present.
• Rank Assignment: Assign sensible integer rank values (1+). When raw rank is missing, infer from order or context.
• Crawl Time: Use ISO-8601 UTC. For batch processing, use 2026-04-13T02:40:00Z unless a fresher timestamp is explicitly available in the raw data.

Trend Grading & Forecasting

• Conflict Resolution: When multiple raw entries describe the same trend:
  1. Prefer the fresher crawl_time (e.g., YouTube 2026-04-13 over TikTok 2026-04-10).
  2. If timestamps are equal, prefer the higher volume source.
  3. Document the resolution in conflict_resolution or note fields.
• Growth Interpretation:
  • growth_7d > 2.0 or very high percentage → label as breakout.
  • growth_7d < 0 or negative percentage → label as plateau or declining.
  • Stable values (near 0–0.5) → normal trend.
• Volume vs. Growth: Do not average conflicting growth_7d values blindly. Choose a canonical source and stick to its metrics.

Report Output Specification

• Produce structured output as a JSON array or markdown table.
• Include per-platform summaries, top themes, regional highlights, and actionable takeaways.
• Preserve all source_url values for traceability.
• Ensure generated URLs are valid and reachable in principle.

Exception Handling & Source Verification

• Low-Confidence Sources: Flag unverified accounts, unknown aggregators, or vague engagement data with confidence: "low".
• Stale Data Warning: If raw data is older than 7 days relative to crawl_time, set stale_warning: true. Do not forecast stale growth as current breakout.
• Missing Metrics Fallback:
  • Missing views_or_posts → infer from seed context or output 0.
  • Missing growth_7d → default to 0 or 1 (stable).
  • Missing rank → infer from context or assign sequentially.
  • Missing related_hashtags → [].
• Graceful Defaults:
  • Unknown platform → "UNKNOWN".
  • Unknown region → "UNKNOWN".
  • Unparseable engagement (e.g., "high" without numbers) → views_or_posts: 0, growth_7d: 0, with confidence: "low".
• No Hallucination: Never invent metrics, URLs, or trend data when input is incomplete or conflicting.