Copy. Paste. Get insights.

Pull all of my videos using list_videos (limit 50, sort by recent) across all platforms. Then perform a statistical significance analysis to identify what variables most strongly predict view count.

For each video, extract the following variables:
- Platform (TikTok, Instagram, YouTube)
- Duration (seconds)
- Day of week posted (derived from post_created_time)
- Hour of day posted (derived from post_created_time, in PT)
- Whether the title contains "part 2", "part 3", or sequel language
- Whether the title contains an artist name (vs. a conceptual/framework video)
- Engagement rate

Then run the following analyses against view count:
1. Pearson correlation + p-value for continuous variables (duration, engagement rate)
2. One-way ANOVA or Mann-Whitney U test for categorical variables (platform, day of week, hour bucket [morning/afternoon/evening], sequel vs. standalone, artist-named vs. conceptual)
3. Flag any result with p < 0.05 as statistically significant
4. Flag any result with p < 0.01 as highly significant

Return results as a ranked table: variable → test used → correlation or effect size → p-value → significance flag. Then write a plain-English summary of the top 3 most actionable insights.

Note: treat Instagram posts with views = 0 as missing data and exclude them from the view count analysis.

Pull my videos using list_videos (limit 50, sort by recent). Rank them by view count and identify the top 5 and bottom 5 performers (exclude any Instagram videos with views = 0). For each of the 10 videos, fetch the full transcript using get_video_analysis.

Then analyze the transcripts across the following dimensions:

Pacing:
- Average words per second (total word count ÷ duration)
- Time to first "hook" statement (first question, bold claim, or pattern interrupt)
- Whether the video opens with context-setting vs. immediate value delivery

Word choice:
- Presence of first-person language ("I", "my", "we") vs. instructional language ("you", "your")
- Use of specificity markers (numbers, percentages, named frameworks, proper nouns)
- Emotional language density (words like "actually", "honestly", "surprisingly", "nobody talks about")
- CTA language presence and position (early, middle, late, or absent)

Return a side-by-side comparison table: top performers vs. bottom performers for each dimension. Then write a plain-English summary of the 3 most actionable patterns separating high from low performers.

Pull all of my videos using list_videos (limit 50, sort by recent). For each video, use the title, description, and any available tags or topics to categorize it into a content theme (e.g., "color theory", "creator tools", "AI", "personal story", "trending audio", etc.). If a video doesn't fit a clear theme, label it "miscellaneous".

Then run the following analysis:
1. Group videos by theme and calculate for each group: average view count, average engagement rate, number of videos posted, view count standard deviation (consistency signal)
2. Identify the top 3 themes by average view count and by total view volume
3. Calculate a "niche concentration score" — what % of my last 50 videos fall into my top 3 themes?
4. Flag any theme with 3+ videos and above-average views as a "validated niche signal"

Return a ranked table: theme → video count → avg views → avg engagement → consistency (std dev) → niche signal flag. Then write a plain-English recommendation for what to double down on and what to sunset.

Analyze the following 3 competitors on [platform]: [username1], [username2], [username3]. For each, pull their top 10 videos using list_creator_videos (sort = top, include_analysis = true).

For each competitor's top videos, extract:
- Content theme / topic category
- Average video duration
- Hook style (question, bold claim, story open, visual hook)
- Whether they use series/part content
- Engagement rate
- View count

Then cross-reference against my own top 10 videos (list_videos, sort = top, include_analysis = true) and identify:
1. Topic categories where competitors are winning but I have zero or minimal coverage ("content gaps")
2. Format patterns where competitors outperform me and I diverge
3. Any topic I cover that competitors don't — potential "owned territory"

Return: a gap matrix (me vs. each competitor, by topic category), a format comparison table, and a plain-English recommendation for 3 content angles I should test.

Pull my top 5 videos by view count using list_videos (sort = top). For each, call list_similar_videos (limit = 8, include_analysis = true) to fetch the videos the algorithm considers most similar to mine.

For each set of similar videos, analyze:
- Average view count of similar videos vs. my video's view count (am I outperforming or underperforming my peer group?)
- Content themes and topics in the similar set — does the algorithm's classification match what I think my video is about?
- Hook styles and formats that dominate the similar set
- Engagement rates in the similar set vs. mine

Then synthesize across all 5 videos:
1. What content buckets does the algorithm consistently place me in?
2. Where am I underperforming relative to my algorithmic peer group?
3. Are there format or hook patterns dominant in my similar videos that I'm not fully utilizing?

Return: a per-video benchmark table (my stats vs. similar video averages), an algorithmic classification summary, and a plain-English answer to: what does the algorithm think I am, and am I winning or losing in that category?