XLSX Spreadsheet Skill
Create, edit, analyze, and convert spreadsheet files with reliable formula checks.
The evidence scopes the skill to spreadsheet processing and requires following user-specified formats, but it lacks explicit user confirmation, data-flow disclosure, backup/rollback, and sensitive-data handling. It permits pip installation when imports fail, while the office helper uses LD_PRELOAD, compiles C code, and may rewrite files in place, so substantial points are deducted. No credential theft, covert exfiltration, or other explicit red-line behavior is shown.
The documentation specifies recalculation, error checking, and several abnormal-case warnings, but the referenced scripts/recalc.py is absent from the supplied evidence. Dependency and LibreOffice availability are not statically proven, and parts of the validator coverage do not fully match the XLSX target. The static ceiling and lack of reproducible tests require a low score.
Trigger conditions, input formats, tool choices, and non-trigger cases are comparatively clear across spreadsheet creation, editing, reading, and conversion. However, non-fit boundaries, Chinese spreadsheet semantics, and cross-environment Excel/LibreOffice compatibility lack evidence; the pip fallback may depend on external sources with unreliable mainland-China reachability, so points are deducted.
The skill is well organized and includes a task table, requirements, formula constraints, gotchas, financial-model conventions, and dependencies. Versioning, changelog, maintenance ownership, update path, and verifiable licensing are incomplete; the SKILL.md claims Proprietary licensing with a complete LICENSE.txt, while the supplied license metadata remains unknown, preventing a high score.
The instructions address common spreadsheet workflows and add useful safeguards for formula recalculation, errors, and maintainable formatting, so the core value is plausible. However, no representative directly usable outputs or execution evidence are supplied, and the key recalculation script is missing from the evidence; correctness and usability can only be partially established under the static cap.
The supplied source is auditable and contains concrete rules, commands, and implementation code. It does not include committed tests, CI results covering the XLSX key paths, or third-party execution records, so formula, recalculation, and file-modification behavior cannot be independently reproduced statically.
- Fallback pip installation introduces supply-chain risk and may depend on network access that is unreliable from mainland China; dependency sources and versions should be pinned.
- recalc.py and its complete runtime environment are absent from the supplied evidence, so recalculation and formula-error checking are not confirmed.
- validate.py --auto-repair and related flows may modify files in place; confirmation and a recoverable backup should be required first.
- The license, version, maintenance ownership, and update path still require verification against the pinned repository contents.
What does this skill do, and when should you use it?
This Agent Skill targets spreadsheet workflows involving XLSX, XLSM, XLTX, CSV, and TSV files. It directs Claude to use openpyxl for formulas and formatting, pandas for bulk tabular data, and markitdown for quick workbook inspection. It also covers recalculation, formula-error checks, financial-model conventions, and safe editing of existing workbooks. It fits teams that need spreadsheet files as deliverables, but the source does not provide a standalone test suite or a general installation procedure for every client.
Reads, creates, edits, and converts XLSX, XLSM, XLTX, CSV, and TSV files; writes formulas, formatting, and comments with openpyxl; imports and exports bulk tabular data with pandas; previews sheets with markitdown; invokes LibreOffice through scripts/recalc.py to recalculate formulas and report errors; and addresses merged cells, macro preservation, cross-sheet references, and formula compatibility.
- A financial analyst builds a forecast model with formulas, input cells, and documented assumptions.
- An operations user cleans malformed CSV or TSV data with misplaced headers and exports a structured workbook.
- An analyst reads multiple Excel files in bulk and consolidates them into a new spreadsheet.
- A user edits an existing XLSX or XLSM file while preserving its formulas, macros, and formatting conventions.
- A team validates formula-containing workbooks with LibreOffice before delivery.
What are this skill's strengths and limitations?
- Covers common spreadsheet workflows including creation, editing, inspection, bulk processing, and conversion.
- Requires formulas instead of hardcoded results and requires formula-error checks before delivery.
- Clearly separates the roles of openpyxl, pandas, and markitdown.
- Provides detailed financial-model conventions for colors, number formats, and structure.
- The skill is marked Proprietary, with complete terms in LICENSE.txt; the repository-level license is listed as unknown.
- Formula recalculation depends on LibreOffice, which may require additional environment setup.
- XLOOKUP, SORT, FILTER, UNIQUE, and SEQUENCE are prohibited, limiting some modern Excel implementations.
- The source provides no standalone test suite, supported-platform matrix, or cross-platform validation results.
- The skill requires a spreadsheet file as the primary deliverable and is unsuitable when the main output is a Word document, HTML report, standalone script, database pipeline, or Google Sheets API integration.
How do you install this skill?
The README documents a Claude Code installation path: run /plugin marketplace add anthropics/skills, choose Browse and install plugins, select anthropic-agent-skills and document-skills, then choose Install now. The documented direct command is /plugin install document-skills@anthropic-agent-skills. The source does not document standalone xlsx installation for other clients.
How do you use this skill?
After loading the skill in a compatible Agent Skills client, submit a spreadsheet task such as: Open downloads/report.xlsx, clean the data, add formulas and formatting, and output the edited workbook. For files containing formulas, follow the skill's validation step: run python scripts/recalc.py output.xlsx [timeout_seconds] and inspect the returned JSON status.