Analyzes Claude projects across cost, efficiency, tool usage, and architecture dimensions using MCP token analyzer tools. Outputs Markdown reports with session rankings, cache rates, subagent overhead, and warnings.
How this skill is triggered — by the user, by Claude, or both
Slash command
/claude-token-analyzer:cta-project-reviewThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Deep analysis of a specific project across cost, efficiency, tool usage, and architecture dimensions.
Deep analysis of a specific project across cost, efficiency, tool usage, and architecture dimensions.
Execute mcp__token-analyzer__sync_db. Skip if already called in this conversation.
If user specifies project_path, use it directly.
Otherwise, execute mcp__token-analyzer__analyze_global and present the project ranking for user to choose.
Execute mcp__token-analyzer__analyze_project with:
project_path: the identified projectinclude_subagents: truesort_by: "cost"limit: 20## 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% |
To calculate overhead, execute mcp__token-analyzer__analyze_project again with include_subagents=false.
Overhead = (cost with subagents) - (cost without subagents).
For top 3 most expensive sessions, execute mcp__token-analyzer__analyze_session with each session_id.
$X.XX USD.125,000).For MCP tool parameter details: ${CLAUDE_PLUGIN_ROOT}/skills/cta/references/tool-reference.md
npx claudepluginhub li195111/claude-token-analyzer --plugin claude-token-analyzerGenerates a quick one-page report summarizing Claude Code token usage, costs, sessions, projects, cache hit rates, subagent ratios, and top spenders. Routes to cost, anomaly, project, or trend audits.
Analyzes Claude Code conversation logs for token usage, costs, cache hit rates, workflow patterns (skills, agents, hooks), and cost optimizations. Generates interactive HTML dashboard.
Analyzes Claude Code session health using PRISM: token usage, CLAUDE.md audits, session failures, and project listings.