This skill is a router. The deep platform knowledge lives in the reference files — this skill tells you when and how to use them.
Cross-Platform Mental Model
Platforms differ on three axes. Knowing where a platform sits on each one shapes every recommendation.
Discovery mechanism:
- Content-graph (TikTok) — the algorithm shows content based on what it is, not who made it. Follower count barely matters.
- Follower-graph (Instagram, LinkedIn) — your existing audience sees your content first. Reach beyond that depends on engagement signals.
- Search + recommended (YouTube) — intent-driven. People find content through search, suggestions, and browse. Evergreen content has a long tail.
- Conversation-graph (X) — content spreads through replies, quotes, and reposts. Real-time relevance matters most.
Content lifespan:
- Ephemeral: Stories (24h), X posts (hours of relevance), TikTok (days to weeks)
- Medium: Instagram Reels/feed (weeks), LinkedIn posts (days to a week)
- Evergreen: YouTube long-form (months to years), YouTube Shorts (weeks)
Audience intent:
- Passive scroll (TikTok, Instagram) — entertainment-seeking, low commitment per piece
- Active search (YouTube, increasingly TikTok) — looking for something specific
- Professional context (LinkedIn) — career-relevant, during work hours
- Conversation (X) — wants to engage, react, discuss
When to Load References
If the conversation involves a specific platform, load references/platforms/{platform}.md from this skill's own directory. If comparing platforms or adapting content across them, load both.
Native vs. Cross-Posted
Every platform's audience can detect cross-posted content. Dead giveaways:
- Wrong aspect ratio or resolution
- Watermarks from other platforms (especially TikTok logo on Reels)
- Caption style that belongs elsewhere (hashtag walls on LinkedIn, professional tone on TikTok)
- Timing and pacing that feels off for the platform
Native content consistently outperforms cross-posted content. When in doubt, adapt.
Algorithm Guidance
Don't give specific algorithm advice without loading the relevant platform reference — algorithms change frequently. Provide the framework (what signals matter, how distribution works) rather than specific numbers. If the reference's last_updated date is more than 6 months old, note that algorithmic details may have shifted.
If the reference can't be loaded, work with the mental model above and be upfront about the limitation.