visualize

# Data Visualization Creator Generate plots/charts/graphs from CSV or JSON via `vl2svg` (Vega-Lite), with column-type detection and chart-type recommendation. ## When to use - User wants a **chart/plot/graph** from tabular data (file or pasted table). - Do **not** use for architecture/sequence diagrams → `diagram`. - Do **not** use for generative art, posters, or non-data images. ## Prerequisites - `python3` and this skill's `main.py` (path: skill directory next to this file). - `vl2svg` on PATH: `npm install -g vega-cli vega-lite`. - Files ≥10MB also need `duckdb` on PATH. ## Procedure Work phases in order. Do not skip. Do not invent column names — only use fields from `analyze` output (or exact user overrides that exist in that output). ### Phase 1 — Materialize data 1. If data is already a file path the user gave, use it as `--data_path`. 2. If the user pasted a table / inline JSON/CSV, write it to a file in the **session scratch directory** (temp). Do not write into the project working directory unless the user explicitly asks to save there. 3. Completion: a real filesystem path exists and is readable. ### Phase 2 — Analyze 1. Run (from skill dir or with absolute path to `main.py`): ```bash python3 <skill-dir>/main.py analyze --data_path <path> ``` 2. Read the JSON stdout. On `"status": "error"`, report `reason` and STOP (or fix path/format and re-run once). 3. Note `recommended_chart_type`, `recommended_x`, `recommended_y`, column `type`...