Agent Token Sentinel

该 Skill 定位为 AI Agent 的成本与配额守护：为单次会话设置美元或令牌硬上限，检测无进展的重复循环并终止进程，并在接近配额前发送高优先级通知。当前仅提供 SKILL.md 说明与一条 openclaw 运行命令，未包含实现脚本、告警渠道及进程终止机制的具体说明。

tmstudio667-commits

热度 72.28

最近更新 2026-05-11 17:28:11

分类：安全依赖：代码执行应用场景：工作流 / 自动化脚本 / 财务

安装技能

复制下面的话术发给你的 Agent，它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill： npx secure-skills-cli add agent-token-sentinel

Metadata

name: agent-token-sentinel
description: Real-time cost & quota guardian for AI Agents. It monitors API usage and automatically kills recursive loops or excessive reasoning to protect your wallet. Optimized for the 2026 Agentic Economy.
metadata: {"openclaw":{"emoji":"💰"},"author":"System Architect Zero","category":"Security"}

Agent Token Sentinel

Protect your treasury from runaway AI logic. This skill acts as a financial circuit breaker, ensuring your Agentic workflows stay within budget by monitoring token burn rates and session quotas in real-time.

Features

Loop Detection: Automatically kills processes that repeat tasks without measurable progress.
Budget Enforcer: Sets hard limits on USD/Token consumption per session.
Alert System: Sends high-priority notifications before you hit your quota.

Usage

npx openclaw skill run agent-token-sentinel --cap 5.00

Architect's Note

In 2026, efficiency is not just an optimization; it's survival.