More screenshots

    Works with the AI Agents you already use

    Claude CodeClaude CodeCursorCursorCodex CLICodex CLIDeepSeekDeepSeekMuseMuse+22 more

    Comment Mining: Voice-of-Customer Research

    1

    Turn public comments into research: sample honestly, classify with a fixed taxonomy, cluster themes with frequency and intensity, and convert every finding into a content, copy, or product decision.

    $5

    Secure checkout via Stripe

    Money-back guarantee

    Converts to your local currency at checkout

    Security scanned
    Comment Mining: Voice-of-Customer Research

    Comment Mining: Voice-of-Customer Research

    Example session with this skill installed

    Our launch video has 12,000 comments. Mine them to fix the landing page and prioritize the next two product tickets.

    • Read your context and instructions
    • Compiled the comment mining

    SAMPLING: Top 50 for messaging language, recent 50 for current issues, controversial 50 for objections — and a random sample of 150 for prevalence. Only the random 150 produce quoted percentages.

    CLASSIFICATION (random 150): Questions 34 (23%), objections 27 (18%), complaints 18 (12%), praise 21 (14%), confusion 15 (10%), requests 12 (8%), buying intent 14 (9%), culture signals 9 (6%).

    THEMES: Largest: "Does it work with [competitor]?" — 18% of the sample, high intensity (repeated, detailed, frustrated). Next: "The price is unclear" — 12%, high intensity. Then: "Love the offline mode" — 9%, medium intensity, in their words. A 6% theme about a minor UI quirk is low intensity and goes to the appendix.

    ACTIONS: The competitor question becomes a compatibility section on the landing page plus a support-team brief. The unclear price becomes a pricing rewrite and a funnel check (buying-intent comments show people wanted to pay and stopped at the price page). The offline-mode praise becomes a headline in the audience's own words. Two complaints — sync failures on one device family — become product tickets, prioritized by frequency times intensity.

    CAVEATS: The report states the sample sizes, names the platform, notes the loud-minority bias, and flags a cluster of twelve near-identical praise comments as suspicious rather than counting them. The findings are directional, and the report says so on page one.

    Connects securely to your tools. The creator never sees your data.

    About this skill

    Nobody fills out your survey, but thousands of people will tell a stranger on the internet exactly what they think. Comments are unfiltered voice-of-customer — messy, biased, and more honest than any focus group. This skill extracts signal without drowning in noise or cherry-picking: sample before you read (top comments for messaging, recent for current issues, controversial for objections — and a random 100–200 for the only prevalence numbers you are allowed to quote), classify with a fixed eight-bucket taxonomy (questions, objections, complaints, praise, confusion, requests, buying intent, culture signals), and keep exact quotes, because paraphrasing kills the value.

    Themes get scored on frequency times intensity — a furious 5% theme outranks a shrugging 15% one — and each theme maps to a decision: repeated questions become FAQs, objections become landing-page sections, complaints become prioritized tickets, praise in their words becomes headlines, buying intent becomes funnel fixes. Themes without an action go to the appendix. Every report carries its caveats explicitly: loud-minority bias, platform skew, negativity bias, astroturfing. A worked example mines 12,000 launch-video comments into a compatibility section, a pricing rewrite, and two product tickets.

    Find more skills like this on the Claude Skills Marketplace, the Codex Skills Marketplace and the Gemini Skills Marketplace, or browse Claude Skills for Social Media.

    • Commenters are the most activated slice, not customers — findings are directional, never census data.
    • Percentages come only from the random sample; under 100 comments there is nothing to quote percentages from.
    • Platform populations differ and are never merged silently.
    • Coordinated or incentivized comments are flagged, not counted — and sophisticated astroturfing can still pass unnoticed.

    How to use

    1. 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. 2

      Sign in to Agensi

      Your agent shows a sign-in link. Sign in and click Allow.

    3. 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 3 days ago

    Permissions

    No special permissions declared or detected

    Listed3 days ago

    What's inside

    Frequently Asked Questions

    Top comments show what the crowd rewarded, which is useful for messaging language — and misleading for prevalence, because the algorithm already filtered. "Most comments say X" from the top 20 is vibes, not research. Prevalence comes from the random sample or it does not get quoted.

    Paraphrasing kills the value. The audience's own phrasing is what goes into headlines and copy — "love the offline mode" in their words beats any paraphrase you write.

    Priority is frequency times intensity. A furious 5% theme outranks a shrugging 15% one — but intensity without frequency is a loud minority, and the rubric exists to stop the loudest voice from winning by default.

    New in Social Media