- Home
- Skills
- Research & Analysis
- Instagram Insights
More screenshots
Works with the AI Agents you already use
Instagram Insights
Finds what works on your Instagram: which theme, hook, length, time and post spacing drive views, retention, shares and follows. Reads Meta Business Suite (no API or token), charts it and gives data-backed tips.
$9.99
Instagram Insights
Example session with this skill installed
Analyze my Instagram Reels from the last 90 days and tell me what's working: which themes, hooks and posting habits get the most views and followers.
- Read your context and instructions
- Compiled the instagram insights
- Generated the document
Summary
I read your Insights → Content table in Meta Business Suite (read-only), tagged each Reel and ran the analysis. Example below: a fictional sample account, @sample_history, 20 Reels from Oct 1 to Nov 8.
6 Reels drove 70% of views, and 5 of them open with “Imagine…”. Both Reels posted minutes after another stalled under 100.
| Metric | Value |
|---|---|
| Views | 35,445 (median 955 per Reel) |
| Accounts reached | 28,820 |
| Hits (≥2× median) | 6 |
| Avg retention | 32% (22 s watched on 66 s videos) |
| Engagement (median) | 10.5% of reach |
| Followers gained | 59 |
What the numbers show
| Finding | Evidence | Confidence |
|---|---|---|
| “Imagine (you in the scene)” hooks perform best | median 3,480 views (5 posts) vs 790 for questions (8 posts) | strong |
| Posting two Reels back to back splits reach | both flops (95 and 70 views) went out 2–3 min after another Reel | weak (2 cases) |
| Theme “Disasters” performs best | median 4,120 views (3 posts) vs 855 for “Ancient world” | medium |
| Longer videos lose retention | length × retention correlation −0.79 | strong |
| Watch time tracks views | watch time × views correlation +0.98 | strong |
Flops (under 25% of the median) are left out of group comparisons because they never got distributed.
What to change
- Open with “Imagine…” + a year. Put the viewer inside the scene in the first sentence.
- Never post two Reels minutes apart. Leave at least a day so each one gets its own test audience.
- Keep it near 60–65 s. Shorter Reels held a larger share of their length.
- Make it shareable. The Emu War Reel alone got 140 of the 473 shares.
Next videos
- Imagine: the night the Hindenburg burned, 1937 (disasters + Imagine is the strongest combination)
- Imagine: London during the Great Smog, 1952
Test for the coming weeks
| Test | How | What to measure |
|---|---|---|
| Same story, two hooks | Post the next disaster Reel with an Imagine hook and a question version as a Trial Reel | Views and avg watch time after 7 days |
Deliverables
report.html: page with KPIs, views per post, median views by theme, hook, era, length, time, weekday and spacing, scatter plots, correlations and a sortable table of every postanalysis.jsonto compare with the next run (--previous)
Next steps
- Review
tags.json(theme, era, hook) and re-run if anything is off. - Apply the top 2 tips for the next 2 weeks.
- Run the analysis again to see what changed.
report.html
PDF · document
Example file from a real run - the skill writes it into your workspace.
Connects securely to your tools. The creator never sees your data.
About this skill
Stop guessing why certain Reels flop while others go viral. This skill reads your Meta Business Suite (via Claude in Chrome, or a CSV export) to extract actual performance data without needing API tokens or passwords. It analyzes themes, hook types, and posting schedules to identify exactly what drives retention and shares on your specific account.
What it does
- Data extraction reads Meta Business Suite insights and account metrics directly from the browser or uploaded CSV exports.
- Content tagging categorizes posts by theme, hook type, and region while matching Reels to local MP4 files to calculate retention.
- Statistical analysis identifies "blowouts" and "flops" based on median view counts and provides confidence levels for every trend found.
- Automated reporting generates a visual Artifact dashboard with performance charts, tips, and data-backed ideas for future videos.
- Trend benchmarking compares current performance against previous rounds to track growth and experiment success.
How it works
- Collect triggers the browser tool to scroll and scrape Meta Business Suite Content Insights or imports a manual CSV export.
- Normalize runs Python scripts to parse raw text into structured JSON and matches video files to post metadata.
- Analyze processes the dataset to find correlations between themes, durations, and engagement metrics.
- Interpret synthesizes the numbers into a playbook of actionable tips and specific video ideas.
- Render publishes a visual report to a Claude Artifact for easy review.
Frameworks & tools
Python 3.12+, Meta Business Suite, Claude in Chrome, and ffprobe for video metadata processing.
Why this beats prompting it yourself
Standard prompting lacks the context of your historical data and can't perform median-based outlier detection. This skill uses local Python scripts to calculate statistical significance, ensuring your strategy is based on math rather than hallucinations.
Use cases
- Identify which content themes consistently achieve 2x the median view count.
- Determine the optimal posting frequency and time of day for your specific audience.
- Audit "flop" videos to find common patterns in hooks or duration that hurt retention.
- Plan next videos and A/B tests based on your own hits.
Known limitations
Automatic collection needs an agent with browser access and a logged-in Meta Business Suite session; otherwise use the CSV export. Measuring retention against video length needs your published MP4 files (optional). With few posts per group, findings are hints to test, not rules. Reports are written in Brazilian Portuguese.
Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Research & Analysis.
Automatic collection needs an agent with browser access and a logged-in Meta Business Suite session; otherwise use the CSV export (recommended for 60+ posts). Confidence is labeled by posts per group: strong (5+), medium (3–4), weak (2). Reports are in Brazilian Portuguese.
How to use
- 1
Connect your agent to the Agensi MCP
Copy the prompt and paste it into Claude, ChatGPT, Cursor or any agent that supports MCP. It sets up the connection for you. Manual setup
- 2
Sign in to Agensi
Your agent shows a sign-in link. Sign in and click Allow.
- 3
Ask for what you need
Your agent loads this skill and uses it on your task. It always asks before buying a skill.
Reviews
No reviews yet
Be one of the first to try it. Every listed skill goes through the trust checks below.
Security scanned
Checked by our 8-point scan before listing
Fresh listing
Recently published to Agensi
Money-back guarantee
Doesn't do what its page promises? You get your money back.
Trust & safety
Security scan
Security scan passed 6 days ago
- Software
Python 3.12+
Git
- Optional
- Google Chromeffprobe