ai-collaboration-notebook AGENTS.md

ai-collaboration-notebook AGENTS.md is an instructions file for Codex, OpenCode from roam-bit/ai-collaboration-notebook. It costs 8,509 tokens per session, scanned A, original, MIT.

A set of instructions for AI coding assistants, based on the ai-collaboration-notebook project, with a quick checklist, collaboration rules, and a log of past mistakes.

In plain words
What is it for?
It is for guiding Claude Code, Cursor, GitHub Copilot, ChatGPT, and similar tools through project work and recording recurring collaboration mistakes.
Why use it?
It gives an assistant a repeatable way to prepare for tasks, choose which rules to read, and learn from earlier errors instead of relying on memory or guesswork.

Instructions file for CodexOpenCode

Install

Getting it into your agent

One page per mod, every tool's command on it. A separate URL per tool would split the same page into five that compete with each other.

agentmods
npx agentmods add instructions/roam-bit/ai-collaboration-notebook/agents-md
Clone the repo
git clone --depth 1 https://github.com/roam-bit/ai-collaboration-notebook

Made for: Codex, OpenCode.

Wrote this? Show the measurements

A badge with what this costs and how it scanned, read live from this page, so it follows the numbers instead of freezing them. Markdown for a README, HTML for a documentation site or a project page.

agentmods badge for ai-collaboration-notebook AGENTS.md

README.md
[![agentmods](https://agentmods.dev/badge/instructions/roam-bit/ai-collaboration-notebook/agents-md.svg)](https://agentmods.dev/instructions/roam-bit/ai-collaboration-notebook/agents-md)
Your own site
<a href="https://agentmods.dev/instructions/roam-bit/ai-collaboration-notebook/agents-md"><img src="https://agentmods.dev/badge/instructions/roam-bit/ai-collaboration-notebook/agents-md.svg" alt="Measured on agentmods" height="20"></a>
Per session 8,509 This file is loaded in full into every session.
When invoked 8,509 The same file — it is already loaded in full.
Security scan A 1 finding. Scan, not verified.
Origin original No closer match found in the catalogue.
Token cost

What it costs to keep this loaded

Counted locally with the o200k_base tokenizer, which is exact for GPT models; Claude uses its own tokenizer and its counts differ. Treat this as one consistent yardstick across the catalogue rather than a bill. Prices are per million input tokens.

ModelPer sessionOnce invoked
Fable 5.1 $0.08509 $0.08509
Opus 5 $0.04254 $0.04254
Sonnet 5 $0.01702 $0.01702
Haiku 4.5 $0.00851 $0.00851

Measured 6d ago against content hash afff38cdd28e, method: parsed. Prices are Anthropic first-party input rates as of 2026-09-05, from the pricing page.

Security

Grade A, and why

ai-collaboration-notebook AGENTS.md scanned grade A with 1 finding against 26 rules in 11 categories — prompt injection, anti-refusal, data exfiltration, privilege escalation, supply chain, agent snooping, system-prompt leakage, SSRF and excessive agency — measured 6d ago.

A static scan of the body, not an audit. Every finding is printed with the line that produced it so you can judge whether it matters here. A mod is markdown that instructs an agent; that is exactly why what it instructs is worth reading.

Makes network callslowCapability

Not a fault in itself. Listed so you know the mod talks to something, and to what.

**真相**:当场 curl 官方 `/models` 实测,现役就是用户说的 V4——我记忆里的旧名是训练截止时代的,厂商早升级了。
AGENTS.md · 548 lines

How it starts

The opening of the file, as written. The whole thing — 548 lines — stays where its author put it; the contents beside it link to each section on GitHub.

AGENTS.md · AI 协作错题本

本文件供 AI 助手直接读取,包含完整的协作协议 + 错题本内容。 适用于 Claude Code、Cursor、GitHub Copilot、ChatGPT 等支持 AGENTS.md 的工具。

项目地址:https://github.com/roam-bit/ai-collaboration-notebook 完整说明见 README.md;本文件 = PROTOCOL.md + MISTAKE-LOG.md 合并版。


这是什么

跟 AI 协作久了你会发现:AI 经常犯同一种错,纠正成本极高,规则越列越多但 AI 选择性遵守。

这是一套已经跑了几个月的协作机制,核心思想:

把 AI 协作当工程问题做反馈循环,不当玄学碰运气。

三层架构(v3):

┌──────────────────────────────────────────────┐
│  CLAUDE.md(自动加载层)                      │
│  内嵌「核心卡」= 30 秒自检 + 17 条规则速查      │
│      + 意图路由表(按任务类型决定深读什么)     │
├──────────────────────────────────────────────┤
│  PROTOCOL.md(规则层)                       │
│  17 条协作规则——按路由表按需深读               │
├──────────────────────────────────────────────┤
│  MISTAKE-LOG.md(观察层,v2 内部分两档)       │
│  活跃层:根因索引 + 聚类视图(开工扫这个)      │
│  归档层:完整推理路径(复盘才读)              │
│  ↑ 同根因 ≥ 3 次 → 升级到规则层               │
│  ↓ 根因休眠 2 次回顾 → 规则退役降级            │
└──────────────────────────────────────────────┘
   ▲ v3 新增:巡检员 agent 定期扫日志补录漏记的失误 + 记量化账本

给 AI 的使用建议:别把本文件全文当"每次开工必读"——把「30 秒自检清单」+ 下面的「17 条规则一行速查」抄进你的自动加载文件(CLAUDE.md / .cursorrules),其余按需回来查。这正是 v2 核心卡机制本身的用法。

17 条规则一行速查1 指令模糊先问别猜 · 2 见次优解主动建议 · 3 技术内容配生活比喻 · 4 回复末尾带 TL;DR · 5 工具产出=待验证 · 6 不迎合反驳必带依据 · 7 事实vs推测分开 · 8 提方案前先搜轮子 · 9 方案选择给对比表+推荐 · 10 "已X"必对应真实工具调用 · 11 责任范围强制兑现 · 12 视觉类需求先要参考 · 13 验收归用户 · 14 持久偏好立刻落盘 · 15 UI完工给截图证据 · 16 说"完工"前过完整验证面 · 17 创意模式强发散先出demo


第一部分:协作协议(PROTOCOL.md)

这是用户(你)跟 AI 助手(Claude / Cursor / Gemini 等)协作的通用规则手册。 跨项目可复用——任何新项目都适用,只需要把这份协议链接进项目的 CLAUDE.md / .cursorrules。


🚨 开工前 30 秒自检清单(每次新任务必跑)

接到任何新任务时,30 秒内回答这 6 个问题,任何一个"不"都要停下补做

  1. 数字加总验证:用户需求里有没有数字(时长 / 字数 / 预算 / 数量 / 范围)?我准备的方案里这些数字加总后等于用户给的总数吗?

  2. 盘点现成资源:这个需求有没有现成轮子?三个来源都查过吗

    • ① GitHub / Hacker News 开源工具
    • ② AI 工具内置的专业 agent / 子代理
    • ③ AI 工具自带能力(FFmpeg / Python / 系统命令等)
  3. 备选方案:我准备给的方案,有 ≥ 2 种备选可对比吗?还是只想到一种就直接给?

  4. 历史教训:上次类似任务踩过什么坑?已升级为硬规则的高频教训:动作 ≠ 结果 / 跨 agent 产出对账 / 推测当事实 / 没搜轮子 / 完工验证不彻底

Read the full file on GitHub · 548 lines

Changes

What this file has done since we first saw it

Hashed on every crawl. A supply-chain change to an agent config is a question of when, not whether, so the history is kept rather than the latest state alone.

  1. 6d ago First seen · 548 lines · 8,509 tokens per session scan A afff38cdd28e

Subscribe to this mod's changes

ai-collaboration-notebook AGENTS.md is an instructions file published in the GitHub repository roam-bit/ai-collaboration-notebook (129 stars, last pushed 2mo ago), licensed MIT. It adds 8,509 tokens to every session, about $0.0425 per session on Opus 5. A static security scan graded it A with 1 finding (makes network calls). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.

Related

Other instructions, from other repositories