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.
npx agentmods add instructions/nexus-research-lab/nexus/agents-mdgit clone --depth 1 https://github.com/nexus-research-lab/nexusWrote 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.
[](https://agentmods.dev/instructions/nexus-research-lab/nexus/agents-md)<a href="https://agentmods.dev/instructions/nexus-research-lab/nexus/agents-md"><img src="https://agentmods.dev/badge/instructions/nexus-research-lab/nexus/agents-md.svg" alt="Measured on agentmods" height="20"></a>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.
| Model | Per session | Once invoked |
|---|---|---|
| Fable 5 | $0.08463 | $0.08463 |
| Opus 5 | $0.04231 | $0.04231 |
| Sonnet 5 | $0.01693 | $0.01693 |
| Haiku 4.5 | $0.00846 | $0.00846 |
Grade A, and why
nexus AGENTS.md scanned grade A with 0 findings 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 today.
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.
Nothing flagged
None of the 26 patterns this scan looks for appear in this file: no shell pipes, no recursive deletes, no credential paths, no hidden text, no instruction-override or anti-refusal phrasing, no agent-config snooping. That is not a guarantee, it is the absence of the things that are checkable.
How it starts
The opening of the file, as written. The whole thing — 103 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
Build & Validation Commands
make dev：同时启动同级 Nexus Control（8020）、Go 后端（8010）和前端（3000）make check-go：默认 Go 门禁，只检查相对上游及当前工作树中发生变化的 Go 包make check-go-fresh：对上述变化包禁用测试结果缓存make check-go-full：显式运行 Go 全量 vet 与无缓存测试，仅用于发布、跨包基础设施变更或用户明确要求make check：运行增量 Go 门禁、前端 lint、前端时间线行为测试、前端 typecheckmake check-backend：Go 后端增量校验，等价于make check-gomake install：执行go mod tidy并安装前端依赖
日常修改先跑目标包测试或 make check-go。Agent 不得默认执行 go test ./... 或 make check-go-full；只有发布、共享协议/基础设施变更、增量范围无法可靠判断，或用户明确要求全量时才执行。
Commit Style
Use English commit messages with an emoji prefix, for example :sparkles: Switch to the Go default runtime path. Keep user-visible changes reflected in CHANGELOG.md.
L1 — 文档地图
代码是机器相，注释是语义相，两相必须同构：任一相变化必须在另一相显现，否则视为未完成。
本仓采用三层分形文档：L1（本节，项目宪法）→ L2（各 Go 包 doc.go 的 L2 头，成员清单 + 暴露接口）→ L3（业务文件顶部 INPUT/OUTPUT/POS 契约）。跨包产品语义只在 docs/specs/ 保留一份当前规范；internal/protocol 类型、nexus.command MCP schema 与 parser 是线格式真相，Skill 只说明模型决策，API Reference 只说明 transport。未来方案、交付计划和未实现字段必须明确标为 non-normative，不能混入当前规范或由多份文档重复定义。
nexus — 用户运行的多 agent 桌面/网页应用；Go 后端 + React web。
技术栈: Go + net/http + WebSocket + SQLite/goose + React19 + Vite + Zustand
<directory>
cmd/ - 可执行入口（nexus-server 服务 + 自动迁移；nexusctl 资源控制 CLI；nexuscfg 配置 CLI；Linux runtime launcher）
web/ - React 前端（features / store / shared / lib，见 web/CLAUDE.md）
desktop/ - macOS AppKit/WKWebView、Windows WPF/WebView2 宿主与 browser-extension（窗口 chrome、bridge、sidecar 生命周期、状态根整体迁移与重启、本机 workspace 文件打开与 macOS 关联应用发现；Windows 用独立原生标题/菜单栏承载全部拖窗与系统命令，WebView 始终保持客户区并通过公开可见性生命周期随主窗口挂起或恢复，Theme/Dialog 将 Nexus token 投影到原生菜单与反馈窗；Chromium 扩展以代次化标签页引用、来源继承租约、round 收尾和增量 AX 快照为 Browser 提供页面、标签页、历史、下载、可见 Agent 指针、交互与用户启用后的完整 CDP 操作）
skills/ - 随产品发布的平台内置 Skill（每个目录自含 SKILL.md、元数据、脚本与按需加载的参考资料）
internal/ - 后端核心（各子包 L2 见其 doc.go）:
protocol/ - 跨 HTTP/WS/前端/运行时的协议真相源（会话/房间/Goal/Execution Graph 与命名工作图模型、NodeRun 历史/可恢复结构化产物/显式 partial/total/控制回连事实与 Room creator/lead 身份、事件、枚举、TS codegen 输入）
runtime/ - nxs/Claude Code 共用宿主主链（bridge client、manager 生命周期、workspace isolation Hook）
service/ - 业务服务（auth 的 Desktop Local 与外部 Control adapter / agent / communication / dm / echo / room / room/realtime / configuration / session / workspace / skills / connectors / automation / llm ...）
service/objectivealignment/ - Goal completion 与 Execution loop guard 共用的无状态目标对齐审计契约
chat/ - 对话领域（dm / room）
handler/ - HTTP / WebSocket 处理器
message/ - runtime/SDK 消息 → Nexus 事件与 assistant 快照的映射投影
echo/ - 用户级 DM 主动跟进策略、attempt 状态与会话覆盖领域模型
automation/ - 定时任务调度域（任务级 capability grant、持久审批、主会话事件派发、run 阻塞与安全恢复）
service/memorymaintenance/ - Nexus 唤醒 nxs 后台记忆维护的宿主协调器
cli/ - nexusctl / nexuscfg 本地命令行装配（按领域文件组织）；模型侧命令不经过 CLI
app/ - HTTP 服务装配与生命周期；server 根包保留进程组合、路由和 worker，goal / execution / workgraph / runtime 子包承载对应功能域适配
mcp/ connectors/ workspace/ - 能力域；mcp 根包持有 physical-round 共用可信上下文与 command receipt，mcp/command 持有 Goal/Execution/Automation 的 `nexus.command` 工具协议和操作适配；宿主自有、与 Nexus 系统功能相关的进程内工具统一挂在单一 `nexus` MCP server 下，各业务包只构建工具定义与固定上下文；模型控制复用内置 Skill，业务输入直接进入宿主，不落临时 JSON；mcp/communication 以 `list_targets` 与上下文感知的 `send_message` 统一 DM、跨会话和当前 Room 通讯，不再设独立 Room MCP 工具包，mcp/browser 通过单个 browser 工具提供完整浏览器操作，mcp/visualize 只暴露 show_widget，skills/visualize 承载生成规范；第三方、用户自定义和 Connector 动态 MCP（包括独立的 `nexus_feishu_docx`）保持各自 server 身份、授权与生命周期，支持原生 MCP 的 Provider 直接挂载自身 server，不提供通用 REST 路由；owner 资源管理复用 nexus-manager / nexusctl，配置管理复用全 Agent 内置 nexus-configuration Skill 与 round-scoped nexuscfg，不再挂载 manager 或 configuration MCP
config/ storage/ infra/ migration/ version/ - 装配、迁移与基础；infra/duework 承载后台 durable work 的合并唤醒、精确 deadline timer 与低频审计，infra/runtimeidentity 承载 Linux UID/GID、ACL、Landlock launcher，infra/confinedfs 承载宿主目录 fd 边界
docs/ - 开源文档入口；README.md 是索引，guides/ 面向用户与作者，images/ 保存图片与导出 SVG，operations/ 面向运维，testing/ 保存维护者回归清单，specs/ 保存当前维护者合同，architecture-html/ 保存可独立打开的图解页面
</directory>
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.
- today Changed · +1 lines · +153 tokens per session e19c6df844a3
- yesterday Changed · +3 lines · +1,048 tokens per session 148b2614fa49
- 5d ago First seen · 99 lines · 7,262 tokens per session scan A c85258456eaa
nexus AGENTS.md is an instructions file published in the GitHub repository nexus-research-lab/nexus (143 stars, last pushed today), licensed Apache-2.0. It adds 8,463 tokens to every session, about $0.0423 per session on Opus 5. A static security scan graded it A with 0 findings. No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
zeroclaw CLAUDE.md
Claude Code instructions for zeroclaw-labs/zeroclaw, covering claude.md — zeroclaw (claude code), claude code settings, hooks and slash commands.
nuwax AGENTS.md
AGENTS.md instructions for nuwax-ai/nuwax, covering ai agent system documentation, 系统概述, ai agent 架构, 核心组件 and ai 功能特性.
AgentEval copilot-instructions.md
Copilot instructions for AgentEvalHQ/AgentEval, covering agenteval - ai coding agent instructions, architecture overview, environment setup, optional: secondary models for comparison and build & test commands.
framework AGENTS.md
AGENTS.md instructions for ai-driven-dev/framework, covering agents.md, behavior, communication, action and memory management.
AgentEval tracing.instructions.md
Guidelines for implementing trace recording and replay.
hands-on-modern-rl AGENTS.md
AGENTS.md instructions for walkinglabs/hands-on-modern-rl, covering repository guidelines, project structure, commands, editing rules and tikz → svg workflow.