Headless Excel Workbook Builder
Generate a deliverable .xlsx file in managed-agent sessions without an open Excel application.
The skill restricts the task to producing a local ./out/<name>.xlsx artifact, declares no network or credential access, and says not to append to an existing workbook, so the scope is relatively contained. It does not specify user confirmation, sensitive financial-data handling, overwrite protection, rollback, or data-flow disclosure, so points are deducted.
The output path, openpyxl dependency, formula conventions, and formatting rules make the happy path plausible. However, there are no dependency installation or version-pinning instructions, input validation, exception handling, diagnostic failure messages, collision handling, or execution tests; static calibration therefore keeps the score at or below 10.
The intended managed-agent/CMA headless scenario is explicit, and the non-fit case of available live Excel tools is stated. Inputs, complex-model boundaries, precise semantic trigger conditions, Chinese-language support, and mainland-China environment fit are not documented, so points are deducted.
The document has a concise structure with name, description, output contract, example, modeling conventions, and a non-use condition. Repository evidence supplies an Apache-2.0 license, official provenance, CI validation, and contribution guidance, but the skill lacks versioning, changelog, named maintenance responsibility, dependency installation guidance, troubleshooting, and FAQs.
The skill directly addresses creating an xlsx artifact headlessly and provides a usable openpyxl example plus input, formula, and checks-tab conventions. There is no representative complete output, formula-result validation, complex-template coverage, or end-to-end execution evidence, so manual review remains necessary and the static score is capped at 7.
The output contract, code example, repository CI, and secret-scan workflows provide limited auditable evidence. The workflows do not cover the skill's key workbook-generation paths, and there is no committed test suite, pinned runtime dependency set, or third-party execution result, so only limited credit is justified.
- The skill instructs the agent to write and run a Python/Bash script but does not define script-permission boundaries or protection against overwriting existing outputs.
- Formulas created by openpyxl may not be calculated or validated; qualified personnel should review inputs, formulas, links, balance checks, and final values.
- Dependency installation, version pinning, abnormal-input handling, and recovery are unspecified, so failures may be difficult to diagnose when openpyxl or write access is unavailable.
- No confidentiality, redaction, or local-retention policy is stated for sensitive financial data.
What it does & when to use it
xlsx-author creates an Excel workbook as a disk artifact for headless managed-agent or CMA sessions instead of editing a live Office workbook. It uses openpyxl through a short Python script and saves the result under ./out/. The skill defines conventions for inputs, formulas, links, named ranges, and balance checks. It fits automated workflows that need a collected workbook file, but it is not a live Excel editing tool.
Writes and runs a Python script using openpyxl to create a workbook; creates ./out/ when needed; saves the workbook as ./out/<name>.xlsx; applies blue, black, and green conventions for hardcoded inputs, formulas, and links; keeps inputs on an Inputs tab; requires calculation cells to use formulas; and adds a Checks tab with TRUE/FALSE model checks such as balance-sheet and cash-flow ties.
- A managed-agent workflow needs to produce and deliver an Excel model without an open Office application.
- A financial analyst wants inputs, calculations, links, and checks organized according to explicit workbook conventions.
- An orchestration layer must collect a generated workbook from a predictable relative path.
- A modeling workflow needs to reduce hardcoded calculation cells and surface balance-check results.
Pros & cons
- Defines a predictable `./out/<name>.xlsx` artifact path for orchestration.
- Provides concrete conventions for inputs, formulas, links, named ranges, and checks.
- Designed specifically for sessions without an open Office application.
- Requires Python and openpyxl, with no version specified.
- Does not edit live Excel workbooks; a different workflow is preferred when live Excel tools are available.
- The supplied source shows no test suite, sample input workbook, or platform-specific validation.
How to install
The README does not document a standalone installation command for this skill. To install the collection in Cowork, open Settings → Plugins → Add plugin, add https://github.com/anthropics/financial-services, and select the relevant agents or verticals. In Claude Code, run claude plugin marketplace add anthropics/financial-services, then claude plugin install financial-analysis@claude-for-financial-services. The skill is located at plugins/agent-plugins/valuation-reviewer/skills/xlsx-author/SKILL.md.
How to use
In a headless managed-agent/CMA session, use a prompt such as: Create a financial workbook as a file artifact, using an Inputs tab for hardcoded values, formulas for calculations, and a Checks tab. Save it as ./out/model.xlsx. The final response should return the relative output path. If live Excel tools named mcp__office__excel_* are available, the source instructs users to use those instead.
Compared to similar skills
The source explicitly distinguishes this file-producing workflow from mcp__office__excel_* live Excel workbook operations; it does not provide enough information for other product comparisons.