going-viral

name	going-viral
description	The viral-content operating loop — research what's already winning, find what actually drives a creator's own hits, ideate against the data, ship one experiment at a time, review, repeat. Use when a SocialGPT user asks "how do I go viral", "why isn't my content taking off", "give me a content strategy", "what should I post next", "how do I grow / get more views / get more reach", or wants a repeatable system instead of post-and-pray. Grounds every step in the user's real data via the SocialGPT MCP and routes into the content-performance-audit, competitor-gap-analysis, and hook-retention-teardown skills for the deep dives. Works without a connection (it still teaches the playbook), but is far stronger with the SocialGPT MCP connected.
license	MIT

Going Viral

Virality is closer to a science than a lottery. The creators who hit consistently aren't luckier — they run a loop: study what already works, form a hypothesis about why, test it with one post, and feed the result back in. This skill is that loop, and it's the conductor for the rest of the SocialGPT toolkit — it grounds each stage in your real data and hands the heavy analysis to the specialist skills.

This skill pairs with the SocialGPT MCP server (https://mcp.gpt.social/mcp). The MCP provides your data; this skill provides the strategy that turns that data into your next post. The other skills below ship the deterministic analysis.

Better with a connection. If the SocialGPT tools aren't available, this skill still works — it teaches the full playbook. But it's far stronger connected: instead of generic advice it studies your real outliers, your drivers, and your hooks. If you haven't connected an account yet, do it here first: https://gpt.social/integrations/mcp (public/competitor analysis works with no account; your own data needs a connected account).

When to use

Trigger on the broad, upstream strategy asks: "how do I go viral?", "why isn't my content taking off?", "give me a content strategy", "what should I post next?", "how do I grow my account / get more reach?", "build me a repeatable system", "I'm posting a lot and nothing's working."

These sit above the three specialist skills — when the user wants the whole system, start here and route into them. When they ask a narrow question ("what drives my views", "compare me to @rival", "teardown my hooks"), go straight to the matching specialist.

The loop

Five stages. At each one, pull the user's real data with the MCP, apply the referenced playbook file, and hand the deterministic analysis to the specialist skill.

Stage	What you do	Pull via the MCP	Hand off to
1. Research outliers	Find creators/formats that go viral repeatedly in or adjacent to the niche; study why.	list_creator_videos(sort="top"), get_creator, list_similar_videos, search_videos; analyze_creator + get_analysis_status for creators not yet in the library	competitor-gap-analysis
2. Find your own drivers	Isolate what actually predicts this account's views — length, timing, platform, format.	list_videos(sort="top"), get_content_profile, get_account_metrics	content-performance-audit
3. Study winning hooks & formats	Read what the top posts' openings and pacing do that the flops don't.	get_video_analysis (transcript, hooks, scenes)	hook-retention-teardown
4. Ideate against the data	Generate many ideas grounded in stages 1–3, then rank and cut hard.	get_content_profile (pillars + voice) + the outputs above	— (reason it out; see references)
5. Ship one + review	Make and ship one post — one clean hypothesis test — then read the result and update what you believe.	get_post_metrics_history, get_growth_summary, get_account_metrics	re-run audit / teardown

The discipline that makes the loop work: one experiment at a time (not a batched calendar), every belief held as a hypothesis until a Gold-vs-Bronze comparison survives it, and a flop diagnosed by stage (bad research vs bad execution) rather than blamed on the idea or the algorithm. See viral-content-model for the full loop and self-analysis-and-iteration for reading a result.

How to go deeper

The references/ folder is the knowledge base — 18 focused playbook files. Load the one that fits the stage you're in rather than reading them all up front:

• The loop & mindset — viral-content-model, viral-mindset-and-myths, generalist-principle
• Research & analysis — reverse-engineering-outliers, viral-performance-drivers, self-analysis-and-iteration
• Formats — proven-viral-formats, viral-format-engineering
• Ideation — content-ideation-pipeline
• Hooks & retention — hook-anatomy, hook-iteration, narrative-tension-and-retention, story-structure-short-form
• Algorithm & distribution — platform-algorithm-fluency, algorithm-seeding-and-trust, early-engagement-diagnostics, communication-algorithm-triple-f
• Execution — execution-craft-and-nuance

Notes

• This skill conducts; it doesn't compute. The hard numbers — significance tests, gap matrices, hook metrics — come from the three specialist skills. Run them at the stages above and feed their findings back into the loop.
• One post, one hypothesis. Resist batching a month of content. You learn nothing between underperformers; you learn everything from one post analyzed deeply.
• Leads, not verdicts. Everything the data surfaces is the best thing to test next, not a guarantee. Hold it loosely and let the next result update it.
• Re-runnable. The loop never ends — re-research and re-audit as the account grows and what works shifts.