Cost & usage report

Read references/mcp.md first for the tool contract and limits. This skill uses only metadata tools — no trace content, so no injection surface. If that file isn't present (this skill was lifted out on its own), ask the user for the plugin's references/mcp.md, or fall back to the hard rules: every tool is read-only, and getcostbreakdown / getusagesummary are the only tools this skill needs.

<!-- copilot:start --> Turn the platform's cost and usage data into a short, decision-ready report.

Workflow

1. Scope. Confirm the date range (getcostbreakdown takes start_date /

end_date) and what the user cares about — total spend, a suspected spike, model mix, or quota headroom. Default to the last 7 days if unspecified.

1. Pull cost. getcostbreakdown. Run it more than once when useful:

group_by: "model" for model mix, "application" for which app spends, "day" for the trend. Use "all" for a first pass.

1. Pull usage/quota. getusagesummary for current billing-period usage and

remaining quota.

1. Report.

- Headline — total spend for the range and where the billing period stands vs. quota. - Top drivers — the models/apps carrying most cost, with shares. - Trend — day-over-day movement; call out any spike and which model/app it came from. - Quota — projected burn vs. remaining (extrapolate linearly from period-to-date usage — state that assumption), and whether the current rate lands over quota before the period ends. - Recommendation (only if the data supports it) — e.g. a high-volume, low- stakes call path that could move to a cheaper model; frame it as a hypothesis to validate, not a certainty.

Keep it tight — a few bullets and one small table, not a wall of numbers.

Never write back to the platform — the server is read-only. <!-- copilot:end -->