Create

Sign in to ReadmeX

or

How we handle your data: Privacy Policy

ReadmeX

Your AI reading companion

What would you like to catch up on?

Pick a briefing, or ask your own question.

Your community and people briefings stay personal to you.

Story

A Practical Guide to AI Agent Context Governance

AI summary

The article presents a practical workflow for governing context in AI agents: before each model call, build a current working view from the effective task, permitted materials, and actual execution records. It separates stable system guidance from dynamic task materials and covers memory admission, versioned tool results and coverage, user corrections, persistence receipts, work progress, and optional context compression or subagents. Its ETF fund-research scenarios, aliases, and code examples are explicitly instructional rather than production interfaces or records of an actual run.

In-depth insight

The guide gives agent builders a practical framework for checking whether instructions, memory, tools, progress, and execution evidence remain valid and correctly represented across handoffs. It also recommends comparing task quality with mechanism costs such as model calls, tokens, latency, and failures.

全景系列 · 原则篇 · 行动篇

一个 Agent 需要知道目标、当前依据、行动边界、历史入口、未知信息的取得路径,以及离完成还差什么。但这些要求不会自己变成一份好的模型输入。

真正落地时,要回答更具体的问题:系统提示词里写什么?记忆在什么时候读?工具返回了一大张表,要不要每轮都带着?用户更新了条件,哪些材料继续有效?模型说“记住了”,我们如何知道真的保存了?

实现的主线,是在每次调用模型之前,从有效任务、获准材料和真实执行记录中构建当前工作视图;行动以后,再让新的观察和变更进入下一轮。 系统提示词提供稳定指导,记忆形成跨任务积累,工具结果与进度支撑当下推进。它们不是一个不断增长的字符串。

下面沿一个教学任务展开:用户比较基金 A、B 的费用与持仓重叠,要求结合上传的个人持仓生成带依据的报告,不执行调仓;研究中途上传新持仓,并明确要求以后基金研究报告不要默认压成短答。场景、输入别名和代码示例均用于教学,不是生产接口或一条实际运行记录。项目经验来自 ETF Agent 的已有实现与文章,不能将全文当作公开教学仓库的完整功能清单。

#ETF Agent#Anthropic

Original · 27 min readRead the original
Group chat

No comments yet. Start the conversation.