Sheetsmith

Pandas-powered CSV & Excel management for quick previews, summaries, filtering, transforming, and format conversions. Use this skill whenever you need to inspect spreadsheet files, compute column-level summaries, apply queries or expressions, or export cleansed data to a new CSV/TSV/XLSX output without rewriting pandas every time.

Security Scan

The skill's code and instructions are consistent with its stated purpose (pandas-based CSV/Excel inspection and light editing); nothing requests unrelated credentials or installs, but be aware it evaluates user-provided pandas expressions (a necessary feature that carries execution risk if you run untrusted expressions) and the README mentions agent-level behaviors (sending files, logging to memory) that are not implemented in the script.

Purpose & Capability

Name/description match the included code and tests: the CLI implements summary/preview/describe/filter/transform/convert for CSV/TSV/XLSX using pandas. No unrelated env vars, binaries, or install steps are requested.

Instruction Scope

SKILL.md and README instruct the agent to run the included Python CLI on files in the workspace or given paths — this is expected. The script uses pandas.DataFrame.query and DataFrame.eval (engine='python'), which necessarily execute user-provided expressions; this is required for the transform/filter features but is a potential execution surface if you run untrusted expressions or pass paths pointing to sensitive files. The README also mentions sending files via Telegram/WhatsApp and updating 'memory' logs — those are agent-level behaviors described in prose but not implemented in the provided code; treat those statements as informational only.

Install Mechanism

No install spec is present (instruction-only install), and the repository includes the script and tests. Dependencies are documented as system packages (pandas, openpyxl, xlrd, tabulate) but no automatic installer downloads arbitrary code during install.

Credentials

The skill requests no environment variables, credentials, or config paths. The operations require only file paths and standard Python libraries; nothing disproportionate is being asked for.

Persistence & Privilege

always is false and the skill does not request or modify other skills or global agent settings. It can overwrite files only when --inplace is explicitly used; no persistent background privileges are requested.

Assessment

This skill appears to do what it says: inspect and modify spreadsheet files using pandas. Before installing/using it: - Review and run the included tests locally to confirm behavior in your environment. - Be cautious when using --query or --expr: these accept pandas expressions and are evaluated (engine='python'), so do not execute expressions from untrusted sources. Preview results before writing and prefer --output (write a copy) instead of --inplace until you're confident. - The README mentions agent behaviors like sending files via Telegram/WhatsApp and updating memory; these are not implemented in the script — treat those as documentation-level suggestions, not automatic behavior of the code. - Verify dependencies (pandas, openpyxl, xlrd, tabulate) are installed in your environment. If you need higher assurance, review the script source (scripts/sheetsmith.py) yourself — it is small and self-contained with no network calls or hidden endpoints.

Overview

Sheetsmith is a lightweight pandas wrapper that keeps the focus on working with CSV/Excel files: previewing, describing, filtering, transforming, and converting them in one place. The CLI lives at skills/sheetsmith/scripts/sheetsmith.py, and it automatically loads any CSV/TSV/Excel file, reports structural metadata, runs pandas expressions, and writes the results back safely.

Quick start

1. Place the spreadsheet (CSV, TSV, or XLS/XLSX) inside the workspace or reference it via a full path.
2. Run python3 skills/sheetsmith/scripts/sheetsmith.py <command> <path> with the command described below.
3. When you modify data, either provide --output new-file to save a copy or pass --inplace to overwrite the source file.
4. Check references/usage.md for extra sample commands and tips.

Commands

summary

Prints row/column counts, dtype breakdowns, columns with missing data, and head/tail previews. Use --rows to control how many rows are shown after the summary and --tail to preview the tail instead of the head.

describe

Runs pandas.DataFrame.describe(include='all') (customizable with --include) so you instantly see numeric statistics, cardinality, and frequency information. Supply --percentiles to add additional percentile lines.

preview

Shows a quick tabulated peek at the first (--rows) or last (--tail) rows so you can sanity-check column order or formatting before taking actions.

filter

Enter a pandas query string via --query (e.g., state == 'CA' and population > 1e6). The command can either print the filtered rows or, when you also pass --output, write the filtered table to a new CSV/TSV/XLSX file. Add --sample to inspect a random subset instead of the entire result.

transform

Compose new columns, rename or drop existing ones, and immediately inspect the resulting table. Provide one or more --expr expressions such as total = quantity * price. Use --rename old:new and --drop column to reshape the table, and persist changes via --output or --inplace. The preview version (without writing) reuses the same --rows/--tail flags as the other commands.

convert

Convert between supported formats (CSV/TSV/Excel). Always specify --output with the desired extension, and the helper will detect the proper writer (Excel uses openpyxl, CSV preserves the comma separator by default, TSV uses tabs). This is the simplest way to normalize data before running other commands.

Workflow rules

• Always keep a copy of the raw file or write to a new path; the script will only overwrite the original when you explicitly demand --inplace.
• Use the same CLI for both exploration (summary, preview, describe) and editing (filter, transform). The --output flag works for filter/transform so you can easily branch results.
• Behind the scenes, the script relies on pandas + tabulate for Markdown previews and supports Excel/CSV/TSV, so ensure those dependencies are present (pandas, openpyxl, xlrd, tabulate are installed via apt on this system).
• Use references/usage.md for extended examples (multi-step cleaning, dataset comparison, expression tips) when the basic command descriptions above are not enough.

References

• Usage guidelines: references/usage.md (contains ready-to-copy commands, expression patterns, and dataset cleanup recipes).

Resources

• GitHub: https://github.com/CrimsonDevil333333/sheetsmith
• ClawHub: https://www.clawhub.ai/skills/sheetsmith