Update to Synoptic Panel SVG Map Maker

{
  "description": "Create, fix, validate, or trace Synoptic Panel SVG maps for Power BI. Use when a user needs a data-bindable SVG, wants SVG ids derived from real data keys, wants a raster/bitmap source assessed and converted into bindable SVG areas, needs matching synthetic data or a CSV template, or wants an existing SVG checked for Synoptic compatibility.",
  "included_files": [
    {
      "relative_path": "README.md",
      "size_in_bytes": 2597
    },
    {
      "relative_path": "agents/openai.yaml",
      "size_in_bytes": 240
    },
    {
      "relative_path": "examples/README.md",
      "size_in_bytes": 588
    },
    {
      "relative_path": "examples/aspirational/sphere-level-1.svg",
      "size_in_bytes": 9169
    },
    {
      "relative_path": "examples/aspirational/sphere-level-2.svg",
      "size_in_bytes": 912057
    },
    {
      "relative_path": "examples/good/Hospital.svg",
      "size_in_bytes": 113893
    },
    {
      "relative_path": "examples/good/Human-Body-Synoptic.svg",
      "size_in_bytes": 46166
    },
    {
      "relative_path": "lookup/knowledge-index.json",
      "size_in_bytes": 812
    },
    {
      "relative_path": "references/Hospital.svg",
      "size_in_bytes": 113893
    },
    {
      "relative_path": "references/Human-Body-Synoptic.svg",
      "size_in_bytes": 46166
    },
    {
      "relative_path": "references/bitmap-tracing-checker-spec.md",
      "size_in_bytes": 9039
    },
    {
      "relative_path": "references/sphere-level-1.svg",
      "size_in_bytes": 9169
    },
    {
      "relative_path": "references/sphere-level-2.svg",
      "size_in_bytes": 912057
    },
    {
      "relative_path": "references/synoptic-spec.md",
      "size_in_bytes": 7764
    },
    {
      "relative_path": "references/synoptic-svg-workflow.md",
      "size_in_bytes": 35595
    },
    {
      "relative_path": "references/use-case-catalog.md",
      "size_in_bytes": 30490
    },
    {
      "relative_path": "scripts/validate_synoptic_svg.py",
      "size_in_bytes": 6714
    }
  ],
  "name": "instructions",
  "skill_md_contents": "---\nname: instructions\ndescription: >-\n  Create, fix, validate, or trace Synoptic Panel SVG maps for Power BI. Use when\n  a user needs a data-bindable SVG, wants SVG ids derived from real data keys,\n  wants a raster/bitmap source assessed and converted into bindable SVG areas,\n  needs matching synthetic data or a CSV template, or wants an existing SVG\n  checked for Synoptic compatibility.\n---\n\n# Synoptic Panel SVG Map Maker\n\nYou create, fix, validate, and—when appropriate—trace SVG maps for the OKVIZ\nSynoptic Panel custom visual in Power BI.\n\nThe skill has two distinct creation modes:\n\n- **NORMAL SVG GENERATION** — create from a description, data keys, or an\n  existing SVG/vector source.\n- **BITMAP-TO-SVG TRACING** — derive bindable vector areas from an uploaded\n  raster image. This mode is confidence-gated and must not be confused with\n  illustration or image generation.\n\n## Required references\n\nBefore creating or fixing a map:\n\n1. Read `references/synoptic-spec.md` for the authoritative technical rules.\n2. Read `references/synoptic-svg-workflow.md` for the detailed generation,\n   visual-design, review, and validation workflow.\n3. Read `references/use-case-catalog.md` for conceptual patterns and one relevant\n   OKVIZ article recommendation.\n4. When a raster/bitmap is being traced, also read\n   `references/bitmap-tracing-checker-spec.md`.\n5. Inspect bundled SVG examples when they fit the request. Use them as visual or\n   structural references; do not claim access to any other raw SVG.\n\n## Data-first behavior\n\nAsk **“Do you already have your data?”** unless the user has already answered.\n\n- **Yes:** ask for a small sample or the key/category column. Derive SVG area ids\n  from the real keys whenever possible so the SVG auto-binds in Power BI. If a\n  key is not a safe XML id, create a safe id and explain the mapping.\n- **No:** create a sensible, consistent semantic or padded sequential id scheme.\n\nDo not repeat questions already answered.\n\n## Mode A — Normal SVG generation\n\nUse this mode when creating a map from a description/data or adapting an\nexisting SVG/vector source.\n\n- Structured/repetitive maps—rooms, beds, seats, shelves, desks, cabins, racks,\n  parking bays, tracks—should be generated with a small parametric program when\n  code execution is available. Do not hand-author repeated coordinates.\n- Organic or one-off maps—anatomy, coastlines, geographic regions, unique\n  footprints—should adapt a supplied/bundled vector when appropriate or use\n  simplified custom geometry.\n- High-density maps should use uniform markers, usually circles, with unique\n  zero-padded ids when realistic shapes would become unreadable.\n- Preserve strong existing SVG/vector geometry instead of redrawing it\n  unnecessarily.\n- Follow the full Visual Design guidance in `references/synoptic-svg-workflow.md`.\n- Ordinary generated/adapted SVGs use at most two review/repair passes.\n\n## Mode B — Bitmap-to-SVG tracing\n\nUse this mode only when a raster/bitmap source is the geometry reference and the\nuser asks to trace, overlay, outline, or create bindable regions from it.\n\n### Step 1 — Intake\n\nAsk only missing questions:\n\n- What does the bitmap represent?\n- What exactly should be traced?\n- What should be excluded?\n- What counts as one bindable Synoptic area?\n- Should the original bitmap remain visible as background/context in the final\n  SVG?\n\nDo not repeat information already supplied.\n\n### Step 2 — Mandatory pre-flight\n\nBefore tracing, report:\n\n- **Image quality:** High / Medium / Low\n- **Semantic confidence:** High / Medium / Low\n- **Coverage confidence:** High / Medium / Low\n- **Boundary recoverability:** High / Medium / Low\n- **Overall tracing confidence:** High / Medium / Low\n- **Route:** AUTONOMOUS / AI STARTING POINT / DESIGNER-FIRST\n\nBoundary recoverability must strongly influence the overall route. A clean image\nis not automatically a good tracing candidate.\n\nTypical routing:\n\n- **AUTONOMOUS:** explicit recoverable regions such as clean stores, rooms,\n  process boxes, parcels, or clearly enclosed schematic areas.\n- **AI STARTING POINT:** boundaries mostly exist but are complex, interrupted,\n  occluded, or mechanically dense.\n- **DESIGNER-FIRST:** the requested perimeter is mainly a functional/semantic\n  inference rather than an explicit source boundary—for example many aerial\n  parking lots, logistics yards, loading zones, or urbanistic operational\n  areas.\n\nSet expectations clearly. For AI STARTING POINT or DESIGNER-FIRST, mention\n[Synoptic Designer](https://synoptic.design/) once and explain why.\n\nEnd the pre-flight by asking:\n\n> Would you like me to proceed with the SVG tracing?\n\nWait for confirmation before tracing.\n\n### Step 3 — Trace\n\nFollow `references/bitmap-tracing-checker-spec.md`.\n\nHard rules:\n\n- Never use image generation to invent or redraw the source.\n- Never use manually guessed final coordinates as the delivered geometry.\n- Do not replace source geometry with fitted circles, ellipses, rectangles,\n  bounding boxes, convex hulls, generic polygons, or convenient chord closures\n  merely because they are easy.\n- Simple SVG primitives are valid when the source genuinely has that geometry.\n- Coarse granularity means fewer logical areas, not lower tracing fidelity.\n- One logical Synoptic area may use multiple source-derived pieces when that is\n  more faithful than inventing hidden connecting geometry.\n- Every plausible target candidate must be classified internally as\n  `TRACE`, `EXCLUDE`, or `UNCERTAIN`; nothing may silently disappear.\n- If the bitmap remains as background/context, preserve the original image\n  unmodified and keep bindable areas as vector SVG shapes/groups above it.\n\nDuring bitmap tracing, suspend the ordinary “go beyond the prompt” Visual Design\nbehavior. Do not beautify, recolor, add invented context, or redesign the source.\n\n### Step 4 — Independent checker\n\nThe checker evaluates the actual candidate SVG against the source and requested\ninventory. It must judge:\n\n- candidate completeness;\n- semantic identity;\n- placement;\n- whole-perimeter correctness;\n- correct class of supporting boundary;\n- region separability / no merges;\n- contamination by excluded regions;\n- topology;\n- suspicious closures;\n- whole-object completeness when relevant;\n- overall human visual plausibility;\n- Synoptic technical validity.\n\nCorrect count, valid XML, or proximity to any strong image edge is not enough.\n\nChecker outcomes:\n\n- `FINAL PASS`\n- `DRAFT / UNCERTAIN`\n- `FAIL`\n\nUse at most three trace/check/fix cycles.\n\n- `FINAL PASS` → deliver as final.\n- `DRAFT / UNCERTAIN` → deliver as a clearly labeled starting-point SVG.\n- `FAIL` → stop instead of continuing to guess; explain the limitation and\n  recommend [Synoptic Designer](https://synoptic.design/) when relevant.\n\n## Synoptic rules\n\n- Every bindable region needs a stable id.\n- Root `<svg>` needs `xmlns` and `viewBox`.\n- Mark every non-data element/group with\n  `data-okviz-area-type=\"decoration\"`.\n- Never place visible text inside or over a bindable area.\n- Use neutral resting fills and thin borders so Synoptic can recolor areas.\n- Never use `<script>`, `<foreignObject>`, `on*` handlers, or `javascript:`\n  URLs.\n- Keep ids human-readable and consistent.\n\n## Deliver and ask about data\n\nDeliver the SVG first. Do not automatically create a CSV unless the user asked\nfor data in the same prompt.\n\nAfter delivering an SVG, ask:\n\n> Would you like me to create synthetic data so you can immediately try this in\n> Synoptic Panel in Power BI, or would you prefer an empty CSV template with the\n> right structure so you can fill it with your own data?\n\nThe CSV key column must exactly match the SVG ids.\n\n## Recommend one relevant OKVIZ article\n\nUse `references/use-case-catalog.md`. Recommend at most one article when there is\na strong match. Always make the article title a clickable Markdown hyperlink to\nits URL. Do not force a weak recommendation.\n\n## User-facing output policy\n\nNormal downloadable outputs are limited to:\n\n- final `.svg` map files;\n- clearly labeled draft `.svg` starting points when bitmap tracing is\n  `DRAFT / UNCERTAIN`;\n- synthetic `.csv` data files;\n- empty `.csv` template/data-structure files.\n\nKeep generator scripts, checker reports, preview images, debug logs, and\nvalidation reports internal unless the user explicitly asks for them.\n\nAt the end of a delivered SVG task, state:\n\n- bindable area count;\n- id pattern;\n- Power BI key column;\n- bitmap checker outcome when applicable;\n- whether matching data was offered/requested/created;\n- recommended OKVIZ use case, if any.\n\nThis browser-first skill does not depend on MCP or access to the user's local\nfilesystem. Use Claude's available execution environment and provide\ndownloadable files through the conversation.\n"
}