# CSV Summary

> MUST be used whenever the user asks to summarize, profile, or get statistics on a CSV file (row/column counts, nulls, ranges, distributions). Always invoke this instead of manually parsing the CSV yourself.

- Skill: `llp42/csv-summary` (Agent Skill)
- Install (CLI): `npx skillmds@latest add llp42/csv-summary`
- Raw SKILL.md: https://api.skillmd.com/api/skills/llp42/csv-summary/raw
- Safety review: pending
- Works with: Claude Code, Claude.ai, OpenAI Codex
- Category: Data & Analytics
- Author: llp42 (https://skillmd.com/u/llp42)
- Updated: 2026-09-22
- Page: https://skillmd.com/skills/llp42/csv-summary

---


Target file: `$1`

1. If `$1` is empty, state that a file path is required and stop.
2. Run `wc -l "$1"` to get the line count.
   - If the file cannot be read or does not look like valid CSV (no consistent delimiter across lines — check with `head -5 "$1"`), say so plainly and stop.
   - If under ~5000 lines, read the file directly with Read and parse it.
   - Otherwise, compute stats with a `python3 -c "..."` one-liner using the standard `csv` module (do not use pandas, do not load all rows into memory/conversation).
3. Compute and report:
   - Total row count (excluding header)
   - Column names
   - Inferred type per column (numeric / text / date / boolean), best-effort from a sample of values
   - Null/empty-value count per column
   - For numeric columns: min, max, mean
   - For columns with under ~20 distinct values: value counts
4. Present the report as a compact table/list. Do not dump raw row data.