scrollmark

Use when analyzing why content performed well or poorly, doing post-mortem analysis, or comparing content performance.

Use when analyzing, evaluating, or generating hooks for short-form video or text posts.

Use when you need to understand platform-specific conventions, algorithm behavior, content formats, or audience expectations for Instagram, TikTok, YouTube, X, or LinkedIn.

Use when interpreting social media content — comments, captions, replies, DMs — and you need to understand tone, subtext, or cultural context beyond the literal words.

Use when adapting content from one platform to another, or advising on multi-platform content strategy.

Use when evaluating whether a trend is relevant, identifying emerging trends, or advising on trend participation timing.

Use when writing in a creator's voice, analyzing a creator's style, or maintaining consistency across content.

Compare a creator's top content against 1-3 named competitors to find where the competitors win and the creator is silent. Use when a SocialGPT user asks "what are my competitors doing that I'm not", "find my content gaps", "what should I steal from @rival", "where am I getting beaten", or wants a side-by-side read of their catalog vs. a competitor's. Pulls the user's top videos and each competitor's top videos via the SocialGPT MCP, extracts themes (from analysis or caption keywords), and writes a visual report of content gaps, owned territory, format gaps, and three angles to test. Requires the SocialGPT MCP server to be connected.

Statistically audit what actually drives a creator's views. Use when a SocialGPT user asks "what drives my views", "why do some of my posts pop", "what should I double down on", or wants a data-backed performance review of their own content. Pulls the user's videos via the SocialGPT MCP, then runs real significance tests (Pearson, Kruskal-Wallis) to rank the factors — length, timing, platform, format — that predict view count, and writes a visual report. Requires the SocialGPT MCP server to be connected.

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.

Tear down what a creator's best hooks do that their flops don't. Use when a SocialGPT user asks "why do my best videos pop and others flop", "what makes a good hook for me", "compare my top vs bottom posts", "what should my openings do", or wants a teardown of their own hooks and pacing. Pulls the user's top and bottom performers via the SocialGPT MCP, reads each one's transcript and hooks, then computes deterministic text/pacing metrics — words-per-second, time-to-hook, first-person vs instructional language, specificity, curiosity words, CTA usage, hook style — and writes a visual report showing the dimensions where winners and losers differ most. Requires the SocialGPT MCP server to be connected.

Drive Showrunner — the AI-powered video generation framework — from an agent session. Use this skill whenever the user asks to create, refine, or export an animated video, explainer, or social clip with Showrunner (or asks for "a video about <topic>"). Covers install/prerequisite checks, the create → inspect → refine → export loop, style/format selection, parsing CLI output, quality self-review against the rubric, and troubleshooting Remotion/FFmpeg failures.