Using a skill in a conversation

No command to remember — just a question

Unlike traditional software, there's no menu or specific command to memorize. You describe your need the way you would to a colleague, and Claude automatically recognizes which installed skill matches, then calls the right Spendesk MCP tools behind the scenes.

The closer your question is to the skill's real use case (visible on its detail page, under "Skill purpose"), the more accurate the answer will be on the first try.

Phrasing examples by role

CFO: "Give me a treasury snapshot across all our entities" or "What's our spend exposure over the last 90 days?" → Claude calls the composite tool spendesk_analyze_spend, best suited for an overview.

AP Manager: "Which invoices are overdue this week?" or "Prioritize the payments to run before Friday" → spendesk_get_due_invoices returns overdue invoices ranked by urgency.

Controller: "Check for duplicate invoices this month" or "Is our GL mapping consistent for this period's spend?" → get_payables combined with get_chart_of_accounts.

Finance Analyst: "Compare spend by cost center over the last 3 months" → spendesk_analyze_spend grouped by cost center.

Procurement: "Which purchase orders don't have a matching invoice yet?" → get_purchase_orders cross-referenced with get_payables.

A conversation, not a one-shot query

The real advantage over a classic Excel export is that you can refine your request as the conversation goes — Claude keeps the context.

This ability to filter, cross-reference, and dig deeper without restating the whole request from scratch is what sets a Claude conversation apart from a plain data export.

What if the answer isn't quite right?

• Rephrase with more business context ("for the France entity only," "excluding credit notes") rather than technical keywords.
• State the time period explicitly (dates, "this month," "the last 90 days") — Spendesk data often spans several years.
• If Claude says it can't fully answer, that's likely a real MCP limitation, not a phrasing issue — see the chapter on Claude as a thinking partner to learn how to tell the difference.

Once you're comfortable with one-off conversations, the next step is automating the questions you ask on a recurring basis: Automating a report with routines.