| name | spreadsheet-data |
| description | Interpret tables and Excel/CSV data—trends, comparisons, anomalies, rollups, and chart suggestions. When users upload spreadsheets, ask what data means, want summaries, or need conclusions visualized. |
Spreadsheet Data
When to use
Before related work, call skill_view{name: "spreadsheet-data"} to load the full body.
Use when drawing conclusions from tables, e.g.:
- User uploads
.xlsx / .csv or pastes a table
- "Look at this data", "what grew fastest", "any anomalies"
- Compare, rank, share, or describe trends
- Narrate trends or comparisons in prose
Not for: pure format conversion, Excel formulas, database ETL (unless user only wants interpretation).
Workflow
- Get data: user attachment, paste, or
read_file on workspace CSV; Excel attachments are parsed to text in chat.
- Confirm columns: if headers are unclear, state assumptions or
clarify key fields.
- Summary before detail: row/column count → key metrics → findings → recommendations.
- Numbers carefully: compute only from visible data; note units and time range; do not invent missing columns.
- Visualization: describe charts in words when asked; use
artifact_save for deliverables (markdown table or notes).
Default output structure
## Data overview
(rows/columns, time range, granularity)
## Main findings
- ...
## Anomalies or watch items
- ...
## Recommendations (optional)
- further analysis / chart type suggestions
Platform conventions
- Default: deliver in chat;
artifact_save for large-table conclusions.
- Do not write one-off table conclusions to long-term memory (background review should not persist them either).
