Local Category Complaint Miner

SKILL: Local Category Complaint Miner
A metro's lowest-rated reviews are a free, honest backlog of unmet needs. Sentiment-filter the angriest Google reviews across many local players, then optionally triangulate against Reddit, X, and Facebook complaints to confirm which pains are real, recurring, and still live — so you can productize or out-execute them.

You are running this skill on API Direct via its MCP tools. Execute the steps below yourself by calling the named tools in order — values in <angle brackets> come from a previous step. Then deliver the result described at the end.

INPUTS:
  - category (required): <category — ASK THE USER>
      The business category to investigate
  - metro (required): <metro — ASK THE USER>
      City or metro to sweep
  - platforms (optional): (optional — e.g. reddit, twitter, facebook)
      Comma-separated list of platforms to run — options: places, reddit, twitter, facebook. Omit to run all of them; name specific platforms to limit the run.
  - country (optional): (optional — e.g. us)
      ISO country code to disambiguate Google Places results for non-US metros. Applied to all Places calls; leave blank to auto-infer from the metro.
  - depth (optional): 3 (default — override if you like)
      How many pages to pull per source (harshest reviews and social posts). Maps to the pages param; higher = deeper, more complete sweep but slower. Defaults to 3.

PLATFORM SELECTION: some steps are gated with "Only run if {platforms} includes X". If the user supplied a `platforms` value, run only the steps whose platform is listed. If the user did NOT supply `platforms`, run ALL steps — every platform the skill supports.

STEPS:
  1. Tool `search_places` — search_places(query="{category} {metro}", pages=5)
     Collect every local provider in the category with its place_id, rating, and review_count; prioritize those with enough reviews to mine. Pass {country} only for non-US metros, otherwise omit it.
  2. Tool `place_reviews` — place_reviews(place_id=<place_id>, sort_by=lowest_ranking, pages=3, get_sentiment=true)
     For each provider pull the harshest reviews and keep only items with negative polarity or anger/disgust as the dominant emotion; capture review dates so you can flag complaints that are still live. 3 defaults to 3.
  3. Tool `place_reviews` — place_reviews(place_id=<place_id>, sort_by=newest, pages=2, get_sentiment=true)
     Cross-check recent reviews to confirm the complaint is still live and not a fixed legacy issue; drop themes that only appear in old reviews.
  4. Tool `search_reddit` — search_reddit(query="{category} {metro}", get_sentiment=true, sort_by=top)
     Only run if {platforms} includes reddit: mine city-subreddit threads and recommendation posts for the same category; keep negative-sentiment comments that name a recurring failure, and treat any theme that ALSO appears in the Google reviews as high-confidence.
  5. Tool `search_twitter` — search_twitter(query="{category} {metro}", get_sentiment=true, pages=3, sort_by=most_recent)
     Only run if {platforms} includes twitter: capture fresh, geo-relevant complaints to confirm the pain is still live right now; keep only negative-sentiment posts and fold them into the same complaint clusters.
  6. Tool `search_facebook_posts` — search_facebook_posts(query="{category} {metro}", get_sentiment=true, pages=3)
     Only run if {platforms} includes facebook: pull local-group and community posts where residents vent about or seek replacements for category providers; keep negative-sentiment items and merge recurring pains into the existing clusters.

DELIVER: A clustered list of recurring unmet needs across the metro's category, ranked by frequency and emotional intensity, with the worst-performing incumbents named. Themes that recur in BOTH Google reviews and the optional Reddit/X/Facebook sweeps are flagged as highest-confidence opportunities.

Note: each underlying tool call is billed at its normal endpoint price; get_sentiment adds a small per-page surcharge. Page through results as needed but stop once you have enough to deliver the outcome.