fleet-auditor

Compare original and translation side by side

🇺🇸

Original

English

🇨🇳

Translation

Chinese

Fleet Auditor: Cross-Platform Agent Token Waste Auditor

Fleet Auditor：跨平台Agent Token浪费审计工具

Detects installed agent systems, collects token usage data, identifies waste patterns, and recommends fixes with dollar savings estimates. Everyone tracks. Nobody coaches. Until now.

Use when: Running multiple agent systems, spending $2-5/day on agents, suspecting idle heartbeats are burning tokens, or want a cross-system cost audit.

检测已安装的Agent系统，收集token使用数据，识别浪费模式，并提供包含资金节省估算的修复建议。此前人人都在追踪数据，但无人提供优化指导。现在，Fleet Auditor来了。

适用场景：运行多个Agent系统，每日在Agent上花费2-5美元，怀疑闲置心跳消耗token，或需要进行跨系统成本审计。

Phase 0: Initialize

阶段0：初始化

Resolve fleet.py path (works for both skill and plugin installs):

bash

FLEET_PY=""
for f in "$HOME/.claude/skills/fleet-auditor/scripts/fleet.py" \
         "$HOME/.claude/plugins/cache"/*/token-optimizer/*/skills/fleet-auditor/scripts/fleet.py; do
  [ -f "$f" ] && FLEET_PY="$f" && break
done
[ -z "$FLEET_PY" ] && { echo "[Error] fleet.py not found. Is Fleet Auditor installed?"; exit 1; }
echo "Using: $FLEET_PY"

Use

$FLEET_PY

for all subsequent fleet.py calls.

Detect systems:

bash

python3 $FLEET_PY detect --json

Parse the JSON output. Report what was found.

If nothing detected, explain: "No agent systems found. Fleet Auditor supports: Claude Code, OpenClaw, NanoClaw, Hermes, OpenCode, IronClaw."

确定fleet.py路径（支持skill和plugin两种安装方式）：

bash

FLEET_PY=""
for f in "$HOME/.claude/skills/fleet-auditor/scripts/fleet.py" \
         "$HOME/.claude/plugins/cache"/*/token-optimizer/*/skills/fleet-auditor/scripts/fleet.py; do
  [ -f "$f" ] && FLEET_PY="$f" && break
done
[ -z "$FLEET_PY" ] && { echo "[Error] fleet.py not found. Is Fleet Auditor installed?"; exit 1; }
echo "Using: $FLEET_PY"

后续所有调用fleet.py时均使用

$FLEET_PY

。

检测系统：

bash

python3 $FLEET_PY detect --json

解析JSON输出，报告检测结果。

若未检测到任何系统，请说明："未找到Agent系统。Fleet Auditor支持以下系统：Claude Code、OpenClaw、NanoClaw、Hermes、OpenCode、IronClaw。"

Phase 1: Scan

阶段1：扫描

Collect token usage data from detected systems:

bash

python3 $FLEET_PY scan --days 30

Report how many runs were collected per system. If this is the first scan, it may take a moment to parse all session files.

从检测到的系统中收集token使用数据：

bash

python3 $FLEET_PY scan --days 30

报告每个系统收集到的运行次数。如果是首次扫描，解析所有会话文件可能需要一些时间。

Phase 2: Audit

阶段2：审计

Run waste pattern detection:

bash

python3 $FLEET_PY audit --json

Parse the JSON output. Present findings ordered by severity and monthly savings.

If no waste found: "Your fleet looks clean. No significant waste patterns detected."

OpenClaw Security + Cost Check (March 2026):

If OpenClaw is detected and version < 2026.3.12: flag as HIGH severity. 31 CVEs including ClawJacked (full agent takeover). Compromised instances run rogue agents that burn tokens undetected. Recommend immediate upgrade.
Check for ClawHub skills matching ClawHavoc campaign patterns (1,184+ malicious skills). These skills exfiltrate data via extra API calls, inflating token spend.
Flag if gateway rate limiting is disabled (CVE-2026-25253). Without it, brute-force attacks can spawn unlimited agent sessions. Recommend:
```
openclaw config set security.rateLimit.enabled true
```

运行浪费模式检测：

bash

python3 $FLEET_PY audit --json

解析JSON输出，按严重程度和月度节省金额排序展示发现的问题。

若未发现浪费："你的Agent集群运行良好，未检测到明显的浪费模式。"

OpenClaw安全+成本检查（2026年3月）：

若检测到OpenClaw且版本低于2026.3.12：标记为高严重级别。存在31个CVE漏洞，包括ClawJacked（完全Agent接管）。被入侵的实例会运行恶意Agent，在未被察觉的情况下消耗token。建议立即升级。
检查ClawHub中的skill是否匹配ClawHavoc攻击活动模式（超过1184个恶意skill）。这些skill会通过额外API调用泄露数据，导致token支出增加。
若网关速率限制已禁用（CVE-2026-25253），标记为问题。若无速率限制，暴力攻击可生成无限Agent会话。建议执行：
```
openclaw config set security.rateLimit.enabled true
```

Phase 3: Present Findings

阶段3：展示审计结果

[Fleet Auditor Results]

SYSTEMS DETECTED
- Claude Code: X runs ($Y.YY)
- OpenClaw: X runs ($Y.YY)

WASTE PATTERNS FOUND
1. [SEVERITY] Description
   Est. savings: $X.XX/month
   Fix: recommendation

2. [SEVERITY] Description
   ...

TOTAL POTENTIAL SAVINGS: $X.XX/month

Ready to act? I can:
1. Show detailed fix snippets for each finding
2. Generate the fleet dashboard for visual analysis
3. Run /token-optimizer for deeper Claude Code optimization

[Fleet Auditor 审计结果]

检测到的系统
- Claude Code：X次运行（$Y.YY）
- OpenClaw：X次运行（$Y.YY）

发现的浪费模式
1. [严重级别] 描述
   预估节省：每月$X.XX
   修复建议：具体方案

2. [严重级别] 描述
   ...

总潜在节省金额：每月$X.XX

是否需要进一步操作？我可以：
1. 展示每个问题的详细修复代码片段
2. 生成集群仪表盘用于可视化分析
3. 运行/token-optimizer进行更深入的Claude Code优化

Phase 4: Dashboard (optional)

阶段4：仪表盘（可选）

If user wants visual analysis:

bash

python3 $FLEET_PY dashboard

This generates

~/.claude/_backups/token-optimizer/fleet-dashboard.html

and opens it in the browser.

若用户需要可视化分析：

bash

python3 $FLEET_PY dashboard

此命令会生成

~/.claude/_backups/token-optimizer/fleet-dashboard.html

并在浏览器中打开。

Phase 5: Deep Dive (optional)

阶段5：深度分析（可选）

For Claude Code specifically, offer

/token-optimizer

for full audit (CLAUDE.md, skills, MCP, hooks, etc.).

For other systems, show the fix snippets from the audit and guide the user through implementing them.

针对Claude Code，可提供/token-optimizer进行全面审计（涵盖CLAUDE.md、skill、MCP、hooks等）。

针对其他系统，展示审计结果中的修复代码片段，并指导用户完成修复操作。

Reference Files

参考文件

Phase	Read
Adapter development	`references/fleet-systems.md`
Detector development	`references/waste-patterns.md`

阶段	参考文档
适配器开发	`references/fleet-systems.md`
检测器开发	`references/waste-patterns.md`

Error Handling

错误处理

No systems detected: Report cleanly, list supported systems
Empty scan results: System detected but no session data in window. Suggest increasing
```
--days
```
Permission errors: Report which files couldn't be read, continue with available data
Corrupted data: Skip bad files, report count of skipped files
fleet.py not found: Check both skill and plugin install paths

未检测到系统：清晰报告，列出支持的系统
扫描结果为空：检测到系统但指定时间范围内无会话数据。建议增加
```
--days
```
参数的值
权限错误：报告无法读取的文件，使用可用数据继续执行
数据损坏：跳过损坏的文件，报告跳过的文件数量
未找到fleet.py：检查skill和plugin两种安装路径

Core Rules

核心规则

Quantify everything in dollars AND tokens
Never read or expose message content (privacy-first)
Report confidence levels alongside findings
Suppress findings below 0.4 confidence threshold
Always show fix snippets with recommendations
Frame savings as monthly recurring, not one-time

所有指标均以美元和token两种单位量化
绝不读取或泄露消息内容（隐私优先）
报告结果时附带置信度
隐藏置信度低于0.4的结果
始终提供修复代码片段和建议
将节省金额表述为月度经常性节省，而非一次性节省