coupler-live-artifactlisted

# Coupler.io live artifact A live artifact is a self-contained HTML page registered with Cowork that persists across sessions and re-fetches data from MCP connectors every time it opens. This skill covers the specific pitfalls of wiring one to a Coupler.io dataflow. Most generic widget guidance (CDN allowlist, sandboxing) lives in Cowork's `create_artifact` tool description — read that first. This skill captures the Coupler-specific traps that cost real time the first time someone hits them. ## When to use The user wants a re-openable view over Coupler.io data: KPIs, funnel, leaderboard, time series, comparison. They have an existing dataflow or are pointing at a data set in their Coupler workspace. The output is a `.html` file registered via `mcp__cowork__create_artifact`. ## Process ### Step 1 — Probe in chat first Before writing a single line of HTML, run these chat-side calls and read the output. All Coupler operations go through the single dispatcher tool `coupler`: `coupler({verb: "call", name: "<operation>", args: {...}})`. Args are snake_case. 1. `search-datasets({query: "<dataflow name>"})` to find the `dataflow_id`, `last_dataset_snapshot_id`, and the dataset `id`. 2. `get-schema({dataset_snapshot_id})` and **read `ai_context` carefully**. It documents column meanings, the funnel order, the SQL conventions, and known caveats. Most pitfalls are flagged here. 3. `get-data({dataset_snapshot_id, query: "SELECT * FROM data LIMIT 3"})` to see the actual row shape.