菜单
This skill should be used when the user asks for a "quick check", "overview", "how much did I spend", "看看狀況", "總覽", or wants a fast one-page summary of Claude Code token usage and costs. The lightest CTA workflow, completes in under 3 minutes. Can also be routed from the main cta skill.
用 Codex 或 Claude 帮你安装 复制这段 Prompt，粘贴到 Codex、Claude 或其他助手里，让它检查 Skill 页面并帮你完成安装。
基于 SOC 职业分类
This skill should be used when the user asks to "analyze token usage", "check costs", "how much did I spend", "token analysis", "session analysis", "cta", or mentions Claude Code usage efficiency, token spending, or cost optimization. Routes to 6 workflow sub-skills: cta-health-check, cta-cost-audit, cta-anomaly-hunt, cta-project-review, cta-trend-watch, cta-usage-pattern. Also handles ambiguous intent with a quick overview. Do NOT trigger for: building token-related features, tokenizer/NLP work, or non-Claude cost analysis.
This skill should be used when the user asks about "使用模式", "pattern 分析", "harness 優化", "工作流建議", "ASCII 圖", or wants to understand how a Claude Code session behaved and what to improve next. Uses classify_session_pattern as the hard-signal source and turns it into actionable workflow guidance.
This skill should be used when the user asks about "anomalies", "problems", "有異常嗎", "哪些會話有問題", "排查", "診斷", or suspects waste or efficiency issues in Claude Code usage. Scans for 6 types of statistical anomalies with severity scoring and automatic drill-down into suspicious sessions. Can also be routed from the main cta skill.
This skill should be used when the user asks about "monthly costs", "cost report", "這個月花多少", "成本報告", "月度審計", "預算", or needs to understand cost distribution by project, model, or time period. Supports cross-month comparison and model cost optimization suggestions. Can also be routed from the main cta skill.
This skill should be used when the user asks to "analyze this project", "project health check", "分析這個專案", "專案健檢", "subagent efficiency", "tool usage analysis", "工具使用", or needs detailed analysis of a specific project's token usage patterns. Provides four-dimension analysis: cost, efficiency, tools, and architecture. Can also be routed from the main cta skill.
This skill should be used when the user asks about "trends", "usage trends", "趨勢", "用量在漲嗎", "預測", "燃率", "burn rate", or wants to understand whether Claude Code usage is increasing or decreasing over time. Supports daily/weekly/monthly granularity with simple forecasting. Can also be routed from the main cta skill.
| name | cta-health-check |
| description | This skill should be used when the user asks for a "quick check", "overview", "how much did I spend", "看看狀況", "總覽", or wants a fast one-page summary of Claude Code token usage and costs. The lightest CTA workflow, completes in under 3 minutes. Can also be routed from the main cta skill. |
One-page summary of Claude Code usage status. The lightest CTA workflow.
Execute mcp__token-analyzer__sync_db. Skip if already called in this conversation.
Execute mcp__token-analyzer__analyze_global with no parameters.
Format results as the following table. Fill every row from the analyze_global response.
## CTA 健檢報告
| 指標 | 值 |
|------|-----|
| 總會話數 | X |
| 總專案數 | X |
| 總成本 | $X.XX USD |
| 平均 Cache 命中率 | X.X% |
| Subagent Token 佔比 | X.X% |
### Top 3 燒錢專案
1. project-name — $X.XX (N sessions)
2. ...
3. ...
### Top 3 最貴會話
1. a1b2c3d4 — $X.XX (project-name)
2. ...
3. ...
After presenting the summary, ask:
「要深入哪個方向？成本 / 異常 / 專案 / 趨勢」
Route the user's choice to the corresponding sub-skill:
| Choice | Invoke |
|---|---|
| 成本 | cta-cost-audit |
| 異常 | cta-anomaly-hunt |
| 專案 | cta-project-review |
| 趨勢 | cta-trend-watch |
$X.XX USD.85.3%).a1b2c3d4).125,000).For MCP tool parameter details: ${CLAUDE_PLUGIN_ROOT}/skills/cta/references/tool-reference.md