Skills for Coding Agents: Claude Code, Cursor and VS Code

Skills for Coding Agents

Once your coding agent is connected to the Xplorr MCP server, these skills tell it how to answer three common questions well: which tools to call, in what order, and what to watch out for. Each one is a small text file you add to your project or your user settings.

Skill	Ask it	Tools it uses
Explain my bill	”Why did our cloud bill go up?”	get_cost_summary, get_mom_comparison, compare_periods, get_top_services, get_top_resources, get_anomalies
Estimate this change	”What will this Terraform change cost?”	estimate_terraform_plan (runs terraform plan and terraform show -json locally, never apply)
Find waste	”Where are we wasting money?”	get_savings_summary, get_recommendations, get_ai_recommendations, get_tagging_violations, list_allocation_rules

All three are read only. Find waste never asks Xplorr to change your infrastructure: it describes the change so the owner can make it.

The commands below download from https://docs.xplorr.io/agent-skills/. Read a file before you add it, as you would any instructions you give an agent.

Claude Code

Skills live in .claude/skills/<name>/SKILL.md in a project, or ~/.claude/skills/<name>/SKILL.md for every project. Claude Code loads a skill when your request matches its description.

for s in explain-my-bill estimate-this-change find-waste; do
  mkdir -p .claude/skills/$s
  curl -fsSL https://docs.xplorr.io/agent-skills/claude-code/$s/SKILL.md -o .claude/skills/$s/SKILL.md
done

Cursor

Project rules live in .cursor/rules as .mdc files. These are set to Apply Intelligently (alwaysApply: false with a description), so the agent uses one when your request matches it, or you can @ mention it.

mkdir -p .cursor/rules
for s in explain-my-bill estimate-this-change find-waste; do
  curl -fsSL https://docs.xplorr.io/agent-skills/cursor/xplorr-$s.mdc -o .cursor/rules/xplorr-$s.mdc
done

VS Code (GitHub Copilot)

Prompt files live in .github/prompts as .prompt.md files and run with /xplorr-explain-my-bill and so on in chat. They run in agent mode with the Xplorr MCP tools (xplorr/*), so name the MCP server xplorr when you add it.

mkdir -p .github/prompts
for s in explain-my-bill estimate-this-change find-waste; do
  curl -fsSL https://docs.xplorr.io/agent-skills/vscode/xplorr-$s.prompt.md -o .github/prompts/xplorr-$s.prompt.md
done

What each skill tells the agent

Explain my bill: total and currency, the largest services and their share, what changed and by how much, and any anomaly that explains it. It never reports a day whose billing is still incomplete (is_final: false) as a drop in spend.

Estimate this change: the monthly change first, then the largest resources, then what Xplorr could not price and why. It states that the figures are AWS on demand list prices without discounts or usage based charges. See Pre Deploy Cost Estimates for what is priced.

Find waste: the five largest savings with the resource, account, reason and monthly saving, the total, then untagged or unallocated spend. It only changes a recommendation’s status when you ask it to.

Related guides

• Hosted MCP to connect Claude Code, Cursor or VS Code
• MCP Tools Reference for every tool and its parameters