Open-source Agent Skills (SKILL.md folders) for review mining, product research, YouTube research, hotel reputation and hiring signals. Each skill tells a coding or research agent (Claude Code, Codex, Cursor, Gemini CLI and others that read SKILL.md) which Apify actor to run, with which input, what the output fields mean and what the run costs.
Disclosure. Arcadium Data built and sells every actor these skills route to (Apify Store user arcadium). The actors are pay-per-event: you pay Apify per result with your own Apify account, at the prices listed in each skill. The skills themselves are free (MIT). No affiliate or referral codes are used in any link.
# all skills
npx skills add arcadiumdata/skills
# one skill
npx skills add arcadiumdata/skills --skill competitor-review-miningOr copy a skill folder into ~/.claude/skills/ (Claude Code) or your agent's skills directory.
/plugin install arcadium-data --marketplace arcadiumdata/skills
On older Claude Code versions: /plugin marketplace add arcadiumdata/skills, then /plugin install arcadium-data@arcadiumdata. The plugin holds the same ten skills, named arcadium-data:<skill>.
- An Apify account and an API token (
APIFY_TOKEN), from Apify Console > Settings > Integrations. - One way to call actors:
- Python 3.9+ only for the two skills that ship a small helper script (standard library only).
| Skill | What it produces | Actors |
|---|---|---|
| competitor-review-mining | Low-star reviews of competitors from Capterra, TrustRadius, Google Play and the App Store, saved in the competitor-profiles/raw/<slug>/<date>/reviews/ layout that the marketingskills competitor-profiling skill reads, plus a ranked complaint list |
arcadium/capterra-reviews, arcadium/trustradius-reviews, arcadium/software-reviews |
| replica-reviews-csv | reviews.csv with source,url,date,rating,text from 3 or more review sites, recent and low-rated first |
arcadium/capterra-reviews, arcadium/trustradius-reviews, arcadium/software-reviews |
| amazon-product-opportunity | Google Trends demand, Amazon competition and 1688 factory prices for a product idea; a drop-in for the Google Trends and supplier steps of product research skills | arcadium/google-trends, arcadium/amazon-products, arcadium/1688-products |
| youtube-transcript-research | Find videos on a topic, pull their transcripts, and summarise claims with timestamps | arcadium/youtube-videos, arcadium/youtube-transcripts |
| hotel-reputation-audit | One hotel (or a hotel and its competitors) across Tripadvisor, Booking.com, Google Hotels, Agoda and Trip.com on one 1-5 scale, with complaint themes and reply rate | arcadium/hotel-reputation, arcadium/tripadvisor-reviews, arcadium/booking-reviews, arcadium/google-hotel-reviews |
| hiring-signals | Which ATS a company uses, how many jobs are open, which teams and countries are hiring, and the open jobs themselves | arcadium/ats-lookup, arcadium/career-site-jobs |
| youtube-transcript | Transcripts of YouTube videos, Shorts, channels or playlists as text, timed segments, SRT or VTT; for when youtube-transcript-api raises RequestBlocked on a server |
arcadium/youtube-transcripts |
| google-trends | Google Trends interest over time, by region, rising queries and Trending now as JSON, more than 5 keywords on one scale; for when pytrends gets HTTP 429 |
arcadium/google-trends |
| trustpilot-reviews | Trustpilot reviews (stars, text, date, company reply) and company rows (TrustScore, star split, reply rate), with star, language and date filters and only-new monitoring | arcadium/trustpilot-reviews |
| arcadium-web-data | A router: from a data request (Amazon, Etsy, 1688, hotel prices and reviews, Yelp, Capterra, TrustRadius, app reviews, Google Trends, Google News, YouTube, career-site jobs, Seek, ATS, AutoScout24) to the actor ID, a first input that works and the price | all 24 public arcadium/* actors |
"Get the transcript of https://www.youtube.com/watch?v=dQw4w9WgXcQ with timestamps." Video links, bare IDs, channels or playlists in one run; segments keep the text, start, duration fields that youtube-transcript-api uses. About $0.026 for 10 transcripts.
"Compare US search interest in notion and obsidian over 12 months." Same parameters as pytrends build_payload, run in the Apify cloud with retries instead of from your IP. About $0.014 for 5 keywords on one scale.
"List the top complaints in the last 200 one- and two-star Trustpilot reviews of bestbuy.com." One row per review plus a free company row; more than 200 reviews per company through filter combinations; onlyNewReviews for a weekly monitor. About $0.17 for 3 companies x 100 reviews.
"Mine 1 and 2 star reviews for Asana, Monday.com and ClickUp and rank the complaints." Capterra and TrustRadius filter by rating on the site; Google Play and App Store reviews are filtered by the helper script scripts/split_reviews.py, which writes competitor-profiles/raw/<slug>/<YYYY-MM-DD>/reviews/<platform>.json and .md. About $0.90 at most for one competitor at 100 reviews per site. For Trustpilot reviews, use the trustpilot-reviews skill.
"Build replica/reviews.csv for Todoist." Writes source,url,date,rating,text with source names capterra, trustradius, google-play, app-store, low-rated rows first, through scripts/to_reviews_csv.py. About $0.45 at most for 200 reviews. This is a data source the user chooses to run; replica-entrepreneur itself asks for reviews collected by reading, and its author does not endorse this skill.
"Is a yoga mat a good product to sell on Amazon US?" Five years of Google Trends interest and rising queries, the Amazon search page (median price, rating counts, sponsored share, bought in past month) and 1688 factory prices at MOQ, then demand, competition and margin scores. About $0.12 per product idea. Replaces the web_fetch Google Trends step in product-research skills.
"What do the top YouTube reviews of the Framework Laptop 13 complain about?" Search in fast listing mode, then transcripts with timed segments, so every claim links to <url>&t=<start>s. About $0.07 for 30 search results and 15 transcripts.
"Audit the online reputation of The Savoy in London." One run across Tripadvisor, Booking.com, Google Hotels (and Agoda, Trip.com) with ratings on one 1-5 scale and a combined summary row (about $1.35 for 300 reviews), or one actor per site for about $0.69 for the same 300 reviews.
"Which of these 40 SaaS companies are hiring sales people in Germany?" ATS detection with open jobs, 30-day velocity, department and country mix for a company list (charged only when a board is found), then the matching jobs and an optional weekly new-jobs watch. About $0.88 at most for 40 companies scanned and 400 jobs.
- The skills are Markdown instructions. They run no code on install, start no server, and use no hooks.
- When a skill is used, it tells the agent to run the user's own Apify CLI (
apify actors call,apify actors info,apify datasets get-items) or the user's own Apify MCP connection. The request goes to the Apify API (api.apify.com) under the user's Apify account and token. It carries the user's inputs: search terms, product or company names, hotel names, and public URLs such as YouTube or Booking.com links. The Apify actors then fetch the public pages and return the rows. - No other destination. The skills send nothing to Arcadium Data or to any third party, use no analytics, and read no credentials other than the user's own Apify login. The
--user-agent arcadium-skills/<skill>flag only labels the CLI call for Apify's usage statistics. - Two skills include a small Python script (standard library only, no network access) that turns the downloaded JSON into files:
competitor-review-mining/scripts/split_reviews.pyandreplica-reviews-csv/scripts/to_reviews_csv.py. - Cost: the user pays Apify per result, at the prices listed in each skill, each with a worked cost example.
- Public data only. The actors do not log in and do not return reviewer names, emails or phone numbers; job rows carry no recruiter names and mask contact details in descriptions. The skills never ask the agent to add personal data from other sources.
- Every skill states the free-plan price per result and the start fee. Paid Apify plans pay less per result.
- Free rows (
error,coverage,summary) are never charged. Filter onrowType.
Every CLI command in these skills uses --json (or --format json), 2>/dev/null and --user-agent arcadium-skills/<skill-name>, the same pattern as apify/awesome-skills.
- Actors on the Apify Store: https://apify.com/arcadium
- Arcadium Data: https://arcadiumdata.io
- Issues and new skill ideas: open an issue in this repo.
MIT, see LICENSE.