agent-ops — Operational Visibility for AI Agents
By Edvisage Global — the agent safety company
Stop flying blind. agent-ops gives you complete visibility into what your agent is doing, what it's costing, and whether it's working.
What This Skill Does
When installed, your agent gains structured operational awareness — tracking its own costs, logging its actions, and generating reports for you, the owner.
Core Capabilities
1. Cost Tracking Protocol
Before every API call, assess and log:
- - Model used: Which model is being called (Opus, Sonnet, Haiku, GPT, etc.)
- Estimated tokens: Input and output token estimate
- Estimated cost: Based on current model pricing
- Task classification: Is this a heartbeat, user request, automated task, or sub-agent spawn?
Maintain a running daily cost log in your memory files:
CODEBLOCK0
2. Action Logging
Every significant action gets logged with:
- - Timestamp: When it happened
- Action type: Post, reply, API call, file read, file write, skill use, tool use
- Target: Where/what was affected
- Outcome: Success, failure, or partial
- Cost: If applicable
Log format:
CODEBLOCK1
3. Daily Summary Report
At the end of each operating day (or on request), generate a summary:
CODEBLOCK2
4. Cost Alerts
Flag when:
- - Daily spend exceeds a threshold (default: $1.00)
- A single action costs more than $0.10
- Heartbeat/maintenance costs exceed 30% of total spend
- Costs are trending higher than the previous 7-day average
5. Model Routing Awareness
When a task could be handled by a cheaper model, note it:
- - Simple lookups → suggest Haiku/Flash tier
- Heartbeats → suggest cheapest available model
- Complex reasoning → confirm current model is appropriate
- Summarization → suggest mid-tier model
How to Use
Tell your agent: "You have the agent-ops skill installed. Start tracking your costs and actions today."
Ask for reports: "Give me your daily operations summary" or "How much have you spent today?"
Limitations (Free Version)
- - Manual report requests only (no scheduled delivery)
- Basic cost tracking (no historical trend analysis)
- Simple action log (no categorized performance metrics)
- No spending limit enforcement
- No multi-agent cost aggregation
Want automated reports, trend analysis, spending limits, and more?
→ Upgrade to agent-ops-pro: https://edvisage.gumroad.com/l/[TBD]
About Edvisage Global
We build practical safety and operations tools for AI agents. Our skills are designed for the OpenClaw ecosystem and install in minutes.
Website: https://edvisageglobal.com/ai-tools
agent-ops — AI代理的操作可见性
作者：Edvisage Global — 代理安全公司
告别盲目操作。agent-ops让您全面了解代理正在做什么、花费多少以及是否正常运行。
该技能的功能
安装后，您的代理将获得结构化的操作意识——跟踪自身成本、记录操作行为，并为您（所有者）生成报告。
核心能力
1. 成本跟踪协议
每次API调用前，评估并记录：
- - 使用的模型：正在调用的模型（Opus、Sonnet、Haiku、GPT等）
- 预估令牌数：输入和输出令牌估算
- 预估成本：基于当前模型定价
- 任务分类：是心跳检测、用户请求、自动化任务还是子代理生成？
在内存文件中维护每日运行成本日志：
每日成本日志 — [日期]
| 时间 | 模型 | 令牌数（输入/输出） | 预估成本 | 任务类型 | 描述 |
|------|------|-------------------|----------|----------|------|
2. 操作日志记录
每个重要操作都会被记录，包含：
- - 时间戳：发生时间
- 操作类型：发布、回复、API调用、文件读取、文件写入、技能使用、工具使用
- 目标：影响的对象/内容
- 结果：成功、失败或部分成功
- 成本：如适用
日志格式：
操作日志 — [日期]
| 时间 | 操作 | 目标 | 结果 | 成本 | 备注 |
|------|------|------|------|------|------|
3. 每日摘要报告
在每个运营日结束时（或按需），生成摘要：
每日运营摘要 — [日期]
成本明细
- - 总预估支出：$X.XX
- 按模型分类：[明细]
- 按任务类型分类：[明细]
- 最高成本操作：[描述]
活动摘要
- - 总操作数：X
- 成功：X | 失败：X
- 创建帖子：X
- 发送回复：X
- 执行API调用：X
警报
4. 成本警报
在以下情况触发标记：
- - 每日支出超过阈值（默认：$1.00）
- 单次操作成本超过$0.10
- 心跳/维护成本超过总支出的30%
- 成本趋势高于过去7天平均值
5. 模型路由感知
当任务可由更便宜的模型处理时，进行标注：
- - 简单查询 → 建议使用Haiku/Flash级别
- 心跳检测 → 建议使用最便宜的可用模型
- 复杂推理 → 确认当前模型是否合适
- 摘要生成 → 建议使用中端模型
使用方法
告诉您的代理：你已安装agent-ops技能。从今天开始跟踪你的成本和操作。
请求报告：给我你的每日运营摘要或你今天花了多少钱？
限制（免费版）
- - 仅支持手动报告请求（无定时投递）
- 基础成本跟踪（无历史趋势分析）
- 简单操作日志（无分类性能指标）
- 无支出限制强制执行
- 无多代理成本汇总
想要自动报告、趋势分析、支出限制等功能？
→ 升级至 agent-ops-pro：https://edvisage.gumroad.com/l/[待定]
关于Edvisage Global
我们为AI代理构建实用的安全和操作工具。我们的技能专为OpenClaw生态系统设计，几分钟即可完成安装。
网站：https://edvisageglobal.com/ai-tools
