| name | cta-project-review |
| description | 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.
|
CTA Project Review — Four-Dimension Project Analysis
Deep analysis of a specific project across cost, efficiency, tool usage, and architecture dimensions.
Workflow
Step 1: Sync Data
Execute mcp__token-analyzer__sync_db. Skip if already called in this conversation.
Step 2: Identify Project
If user specifies project_path, use it directly.
Otherwise, execute mcp__token-analyzer__analyze_global and present the project ranking for user to choose.
Step 3: Analyze Project
Execute mcp__token-analyzer__analyze_project with:
project_path: the identified project
include_subagents: true
sort_by: "cost"
limit: 20
Step 4: Output Four-Dimension Report
## CTA 專案分析 — project-name
### 成本面
| 指標 | 值 |
|------|-----|
| 總會話數 | N |
| 總成本 | $X.XX USD |
| 平均每會話 | $X.XX |
| Top 3 最貴會話 | a1b2c3d4($X), e5f6g7h8($X), ... |
### 效率面
| 指標 | 值 |
|------|-----|
| 平均 Cache 命中率 | X.X% |
| 低 cache 會話比例 | X.X% |
### 工具面
| 工具 | 總調用 | 使用會話數 | 每會話平均 |
|------|--------|-----------|-----------|
| Read | X | N | X.X |
| Bash | X | N | X.X |
| Agent | X | N | X.X |
### 架構面
| 指標 | 值 |
|------|-----|
| Main 會話數 | N |
| Subagent 會話數 | N |
| Subagent Token 佔比 | X.X% |
| Subagent Overhead | $X.XX (估算) |
| 模型分布 | Opus X% / Sonnet X% / Haiku X% |
Step 5 (Optional): Subagent Overhead
To calculate overhead, execute mcp__token-analyzer__analyze_project again with include_subagents=false.
Overhead = (cost with subagents) - (cost without subagents).
Step 6 (Optional): Session Drill-Down
For top 3 most expensive sessions, execute mcp__token-analyzer__analyze_session with each session_id.
Behavior Rules
- If no project_path given, show project ranking from analyze_global and let user choose.
- Calculate "per session average" as total_invocations / session_count for tool ranking.
- High Agent tool usage often correlates with high subagent ratio — cross-validate architecture dimension.
- Cache hit rate <70%: mark with warning. Subagent ratio >20%: mark with notice.
Output Rules
- Use 繁體中文 for prose, English for technical terms.
- Currency:
$X.XX USD.
- Percentages: one decimal place.
- session_id: first 8 characters only.
- Token counts: thousands separator (
125,000).
Additional Resources
For MCP tool parameter details: ${CLAUDE_PLUGIN_ROOT}/skills/cta/references/tool-reference.md