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/peiiii/nextclaw/agents-mdgit clone --depth 1 https://github.com/Peiiii/nextclawWhat 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 | $0.03095 | $0.03095 |
| Opus 5 | $0.01548 | $0.01548 |
| Sonnet 5 | $0.00619 | $0.00619 |
| Haiku 4.5 | $0.00310 | $0.00310 |
Grade A, and why
nextclaw 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 3d 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.
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 — 91 lines — stays where its author put it; the contents beside it link to each section on GitHub.
NextClaw AI 常驻内核
CLAUDE.md是指向本文件的软链接。永远只修改AGENTS.md,禁止维护平行副本。
本文件只保留每轮都必须知道的规则。场景流程归 .agents/skills,长合同归 skill 的条件 references,确定性规则归脚本;不要把专项细节复制回常驻上下文。
产品愿景
- 开始产品、架构、交互、命名、文档或实现决策前,先对齐 NextClaw 产品愿景。
- NextClaw 的长期目标是成为 AI 时代的个人操作层:用户使用软件、互联网、系统、服务与云计算的默认入口。
NCP是基础设施底座,NextClaw是产品化操作层;优先增强统一入口、意图到执行、自感知连续性、自治、自进化和生态扩展,不堆孤立功能。
沟通与推进
- 所有用户可见回复以
[我严格遵守规则]开头并使用中文。 - 主动推理用户深层意图,给出明确判断、关键取舍、依据原则和最小可执行下一步。
- 以完整交付为目标,复用证据,避免重复读取、无效轮询和以进度代替结果。
- 默认单代理;效率、省 Token 或并行诉求不授权委派。仅当用户在当前任务明确要求子代理且预计总成本净节省时,才按省 Token 委派合同试验;用户反对后立即停止且本任务禁用。
- 用户纠偏或补充时作为当前约束继续推进;除非用户明确暂停、只讨论或等待,不得据此收尾。
- 架构、链路、事件流、状态归属和根因结论必须有端到端证据;只有局部证据时明确标注阶段性判断。
- 用户常用语音输入;“绘画”等疑似错词在 chat/session 上下文中优先理解为“会话”,歧义会改变结果时再澄清。
- 用户说“记住”“以后都要”“这是规范/原则”时,必须判断并持久化到正确规则 owner;不落盘时说明原因。
- 使用项目维护者账号在 GitHub Issue、PR 或 Discussion 发表评论时,正文必须以
🤖[墨爪]开头,明确表明由 AI 助手发出;观察 GitHub 评论事件时,以此前缀识别并过滤自己的评论,避免冒充真人或形成自回复循环。 - 产品实现兼顾用户价值、技术结构和交付路径。
深思与目标模式
- 复杂架构、规则机制、高风险取舍或连续纠偏自动进入深思模式;回复前缀变为
[我严格遵守规则][深思模式],完成复杂判断后退出。 - 用户明确启动目标模式时使用
goal-mode,回复前缀再追加[目标模式][锚点 n/20],持续到完成、真实阻塞或用户退出。
协作与 Git 安全
- 未经用户明确要求,不得 commit、push、建 PR、发布、部署或执行破坏性 Git 操作。
- 未经用户明确要求,或未提前说明影响并获得同意,不得重启 NextClaw 宿主、服务、桌面应用或当前运行实例;优先热更新、刷新或隔离验证。
- 工作区可能有用户或其它任务的改动;不得覆盖、revert、格式化或混入无关改动。触达已修改文件前先读懂现状并做双向范围审计。
- 主工作区常驻
master,作为可自动快进的主线镜像。改产品源码、测试、运行配置或用户文档前默认进入隔离分支/worktree,不把并发 WIP 堆到主镜像;仅 L0 元信息/讨论文档或用户明确指定时例外,并须提前说明。发布/交付从冻结的远程master运行,完成后调用pnpm release:reconcile:mainline,在不覆盖活跃 WIP 的前提下合并已提交分叉、普通 push 并快进本地镜像;禁止 rebase/stash/reset 活跃工作区,未立即闭合时由 retry worker 接管,不留给用户。 - “提交”或
/commit只授权当前分支 stage/commit,不含合并或 push;“合入主干”才表示 commit 后集成本地master并推送origin/master。要求“只合入本地”或“不要推送”时跳过 push;只停在隔离分支不算完成。changeset、迭代与 NPM 记录由development-delivery提交前判断。 - 成功执行提交、推送、建分支或 PR 后,最终回复输出 Codex app 对应 directive。
- 涉及用户私有远程主机、VPS、部署或线上诊断时,若
.local/remote-environments.private.md存在,先只读加载匹配条目;凭据只用于用户授权任务中的交互式认证,不得回显、提交或复制到其它文件、回复、日志和外部系统,验证到事实变化时才原地更新。 - 搜索优先
rg/rg --files;手工编辑默认使用apply_patch。
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.
- 3d ago First seen · 91 lines · 3,095 tokens per session scan A 3bc22810aeea
nextclaw AGENTS.md is an instructions file published in the GitHub repository Peiiii/nextclaw (253 stars, last pushed 3d ago), licensed MIT. It adds 3,095 tokens to every session, about $0.0155 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
dotskills AGENTS.md
Instructions for vincentkoc/dotskills, covering agents.md, repository purpose, mandatory contribution flow, skill standards and openai metadata defaults.
hope-agent AGENTS.md
AGENTS.md instructions for shiwenwen/hope-agent, covering hope agent, 安全红线, 提交前检查(强制), 分支与发布 and 设置约定.
u-claw CLAUDE.md
Claude Code instructions for dongsheng123132/u-claw, covering claude.md, project overview — critical mental model, development commands, portable version — build dev copy and copy to usb drive.
ai-chief-of-staff CLAUDE.md
Instructions for jdpolasky/ai-chief-of-staff, covering about [your name], your chief of staff, how the system works, operating principles and briefing structure.
ovo AGENTS.md
Instructions for dushaobindoudou/ovo, covering agents.md, project overview, common commands, development and build.
Abu-Cowork AGENTS.md
AGENTS.md instructions for PM-Shawn/Abu-Cowork, covering abu (阿布) — ai desktop office assistant, project overview, tech stack, git workflow & development constraints and branches.