Update to Popety.io

{
  "description": "Hunt under-exploited parcels with development potential in a commune or canton. Use when the user asks about development potential, under-exploited parcels, site sourcing, densification opportunities.",
  "included_files": [
    {
      "relative_path": "references/popety-kit.html",
      "size_in_bytes": 17333
    }
  ],
  "name": "find-development-sites",
  "skill_md_contents": "---\nname: find-development-sites\ndescription: \"Hunt under-exploited parcels with development potential in a commune or canton. Use when the user asks about development potential, under-exploited parcels, site sourcing, densification opportunities.\"\ndisable-model-invocation: false\n---\n\n<!-- GENERATED FROM api/popety-mcp/src/prompts.ts — do not edit; run pnpm --filter @popety-io/mcp generate:integrations -->\n\n# Find development sites\n\nHunt under-exploited parcels with development potential in a commune or canton.\n\n**Arguments to collect from the user:**\n\n- `location` — Commune (municipality) name, e.g. \"Pully\", or a 2-letter canton code, e.g. \"VD\". (required)\n- `min_area_m2` — Minimum parcel area in m² (optional). (optional)\n\n## Workflow\n\nFind under-exploited parcels with development potential in <commune or canton>. Follow these steps exactly:\n\n1. PREVIEW THE SEARCH (free)\n   Call search_entities with:\n   { entity_type: \"lands\", filters: { municipality: \"<commune or canton>\", under_exploited_score_min: 60, no_building_zone: \"hide\" },\n     sort: \"under_exploited_score_desc\", preview: true, limit: 10 }\n   (no_building_zone: \"hide\" keeps only constructible parcels — never shortlist land in a no-building zone. Preview with the SAME limit you will commit with: object_count/cost are quoted at that limit, matching_total is the full count.)\n   Note the object_count and cost — this search bills per returned object.\n\n2. RUN THE FULL SEARCH\n   Repeat the same call without preview, passing the query_id from step 1 and limit: 10.\n   Extract per parcel: land_id, code_number, municipality, area, under_exploited_score, development_score.\n\n3. CHECK ZONING HEADROOM FOR THE TOP 3\n   For the 3 highest-scoring parcels, call get_land with { land_id: <land_id>, include: [\"zoning\"] }.\n   Note: zone name (main_lupa_name), authorized IUS (ius_authorized) vs what is currently built — the gap is the densification headroom.\n\n4. THE DOCUMENT (blueprint — follow the kit)\n   Ranked parcel CARDS (kit card primitive): per parcel code_number + land_id, commune, area m², the two scores as GAUGES (under-exploited, development), zone + IUS, and one line on WHY it is under-exploited (built density vs authorized IUS — estimate the headroom in m² where the indices allow). Open with a verdict naming the one or two credible targets and why the rest rank lower.\n   Notes-line material: under-exploited and development scores are Popety proprietary indicators, not building permits. Next step: run the feasibility_snapshot prompt on the best candidate to check zoning rules, restrictions, and hazards before acting.\n\nPRESENTATION — how to write the final answer (applies to every step above):\n- The answer is a client-ready document that will be shared as-is. Not a chat log, not an analysis diary.\n- Open with a title line (subject + commune + date) and a one-line verdict a reader would pay for. Never open with method, tool narration, or what you are about to do.\n- The reader is a real-estate professional (broker, investor, developer). Every section closes with ONE analyst paragraph stating the professional implication — pricing, absorption, sourcing, underwriting — never a definition or a tutorial sentence. Show sample sizes and visibly de-emphasise figures resting on fewer than 3 observations.\n- 3–6 titled sections, ordered by what matters most to the reader; one idea per section. Prefer short prose with embedded figures; use a table only when comparing 3+ items across 2+ dimensions, max ~6 rows, one comparison per table. Select the figures that change the reader's decision — do not dump every number you retrieved.\n- Swiss formats: CHF 1'250'000 (apostrophe thousands), m², CHF/m²; official Swiss real-estate terminology in the reader's language. Write the whole document in the language the user wrote in.\n- Caveats: at most ONE short \"Notes\" line at the end of the document (e.g. coverage gaps, asking ≠ realised prices). Never inline a disclaimer after a figure. Never mention tools, credits, section ids, API mechanics, or observations \"for the platform team\" inside the document — if you have a genuine data/product observation, put it after the document under a separate \"---\" divider, in one or two lines.\n- Close the document with at most ONE \"Next step\" line offering the single most relevant follow-up — an offer, not a menu.\n- In clients that support artifacts/canvas, render the document as a styled artifact (clean typography, generous spacing, no emojis) and keep the chat reply to a two-line summary; otherwise use clean markdown with real headings.\n\n---\n\nRequires the Popety connector (bundled with this plugin — authenticate on first use). Follow the workflow's cost gates: free/preview steps before any paid call.\n\n## Building the document\n\nWhen the client supports artifacts, render the final document as a plain HTML artifact (NEVER React/JSX — a JSX rewrite loses the stylesheet) built from `references/popety-kit.html` (bundled with this skill): read the kit FIRST, paste its ENTIRE <style> block verbatim as the stylesheet (never write your own CSS), give every SVG mark its literal hex fill/stroke attribute PLUS the kit mark class (mk1-mk4, mkband, mktick, mkline), and follow its primitives, helpers and rules — Popety brand in both themes, Swiss number formats, inline-SVG charts with tooltips, one analyst paragraph per section written for a real-estate professional. The finished document must visibly carry the Popety cyan (brandbar, kicker, verdict block, KPI accents, chart fills) — a grey/monochrome document means the kit was not applied. Without artifact support, apply the same structure in clean markdown.\n"
}