Borrowing it
Nothing to install: this file belongs to D-Robotics/moss. Take a copy, put it at the same path in your own repository, and replace the rules that are about this project with yours.
curl -O https://raw.githubusercontent.com/D-Robotics/moss/main/AGENTS.mdgit clone --depth 1 https://github.com/D-Robotics/mossWrote 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/d-robotics/moss/agents-md)<a href="https://agentmods.dev/instructions/d-robotics/moss/agents-md"><img src="https://agentmods.dev/badge/instructions/d-robotics/moss/agents-md/github.svg" alt="Measured on agentmods" height="20"></a>Or the 80×15 button, for a site that already has a row of RSS and ATOM ones. Only the verdict fits; the numbers stay here.
<a href="https://agentmods.dev/instructions/d-robotics/moss/agents-md"><img src="https://agentmods.dev/badge/instructions/d-robotics/moss/agents-md.svg" alt="Reviewed on agentmods" width="80" 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 5.1 | $0.05471 | $0.05471 |
| Opus 5.5 | $0.02188 | $0.02188 |
| Sonnet 5.5 | $0.01094 | $0.01094 |
| Haiku 4.5 | $0.00547 | $0.00547 |
Grade C, and why
moss AGENTS.md scanned grade C with 2 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 12d 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.
Reaches for credential fileshighPrivilege escalation
SSH keys, cloud credentials, git-credentials, .npmrc, /etc/shadow: reading these is how a config file becomes a credential leak.
与敏感读取档（`/etc/shadow`、私钥、`sshd_config`、`authorized_keys`）在 TTY 上确认，headless 拒绝， Runs shell commandslowCapability
Expected in a hook, worth knowing in a rule or an instructions file.
`runProcessSync`），工具执行路径禁用 `execFileSync` / `execSync`。 How it starts
The opening of the file, as written. The whole thing — 205 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AGENTS.md
本文件是 Moss 仓库对所有 coding agent 的项目指令（被 git 跟踪、可审查）。
仓库身份
Moss 是一个精简的跨平台 coding agent harness：TypeScript / ESM 单包仓库（根 package.json，
包名 @rdk-moss/agent，bin 为 dist/cli.js），Node ≥ 22.16.0，运行于 Linux / macOS / Windows。
核心能力（也是唯一应当存在的范围）：agent loop、工具框架（src/tools/）、上下文管理
（src/context/）、provider（src/provider/）、安全（src/safety/）、会话
（src/core/session/）、子代理（src/core/subagent/）与 CLI（src/cli/、src/cli-main.ts）、
设备抽象（src/device/ + src/contracts/device.ts + src/tools/device-tools.ts，SSH 连 RDK /
Linux 真机，目标是机器人闭环 Goal→…→Deploy→Verify→Repair→Physical Acceptance）。
共享契约在 src/contracts/。
范围变更（2026-09-30 v0.14–v0.20 路线图冻结决策，见
docs/superpowers/plans/2026-09-30-moss-v014-v020-roadmap.md；同日机器人闭环总指令为
更高优先级的产品方向，两线共存）：
- 解冻：MCP 客户端（仅客户端；stdio + streamable HTTP 双传输、工具懒加载、
src/core/mcp/，v0.16）与轻量 skills（SKILL.md 渐进披露、src/core/skills/，v0.16）。 - 翻案：v0.17 起引入可选全屏 TUI（ink，严格圈禁
src/cli/tui/动态 import， 无 TTY 或--no-tty回退 readline REPL；REPL 永久保留）。 - 维持冻结：memory（跨会话自动记忆）/ mesh / observability / orchestration / web-ui / 插件市场。不要重新引入这些子系统。
代码规范（必须遵守）
- Prettier 负责格式；TypeScript/JavaScript/MJS 文件名用 kebab-case；Node 内置模块用
node:前缀；仅类型导入用import type；ESM 相对导入带.js后缀。 - 禁止
any；不要留下未处理的 Promise；跨工具 / provider / CLI 边界的错误转换为MossError（src/errors.ts）并保留原始 cause；禁止catch (err: any)。 - 所有子进程必须经
src/utils/run-process.ts（runProcess/spawnProcess/runProcessSync），工具执行路径禁用execFileSync/execSync。 - 新工具必须声明 side-effect 元数据（readonly vs mutating 驱动审批）。
- 非流式 LLM provider 必须声明
capabilities: { streaming: false }。 - 面向用户的成功消息必须来自真实结果（probe / exit code / post-condition），不得是固定字符串。
- 凭据只从
.env或环境变量读；不硬编码、不写日志、不传给外部服务。
常用命令
| 命令 | 用途 |
|---|---|
npm run build |
清理并构建到 dist/ |
npm run typecheck |
全量类型检查 |
npm run lint / lint:fix |
ESLint（0 warning） |
npm run test |
构建 + 运行 test/*.spec.mjs（面向 dist/） |
npm run test:filter -- --filter <name> |
只跑匹配的 spec（至少匹配 1 个，否则失败） |
npm run smoke |
CLI 冒烟（版本 / 帮助 / PTY 启动） |
npm run check |
format:check + lint + typecheck |
npm run verify |
check + test + smoke，交付前必须绿 |
npm run bench [-- --task <id> --samples <n>] |
agent 能力基准（bench/tasks/，DeepSeek 基准模型，结果落 bench/results/，不入库） |
npm run bench:ab -- <engine> |
hard 层 A/B 对照（best-of-n / reasoning-high / model-routing，--samples <n> 可调），输出默认开/关建议 |
npm run bench:noise -- <label1> <label2> [...] |
同 SHA 重复跑聚合成噪声带（bench/results/noise-band.json） |
npm run bench:swe -- [--samples N --concurrency K --label L --filter s --eval] |
SWE-bench Verified 100 实例锁子集（bench/boards/swebench-instances.json）：容器内 moss headless 产 patch + 官方 swebench harness 判分；密钥经 MOSS_BENCH_API_KEY |
npm run bench:deepswe -- --tasks <dir> [--samples 8] |
DeepSWE v1.1 同模型 harness 对比（经 Pier；已发布分数在 bench/boards/deepswe-v1.1-harness.json；默认 1 样本，发布口径 8 样本、temperature 1.0、top_p 0.95） |
npm run bench:device -- --dry | --target sim | --target real |
设备任务成功率（bench/device-tasks/）；PASS = Moss 裁决通过且独立探测与证据一致。见 docs/bench/device-bench.md |
MOSS_REAL_TERMINALS=1 npm run test:filter -- --filter tui-real-terminals |
P7 真实终端核对（tmux / GNU screen / Terminal.app，先 build）；默认 npm test 不跑，缺终端则 skip。见 docs/cli-parity/tui-real-terminals.md |
npm run bench:tui-feel |
TUI 体感基准（scripts/tui-feel/，PTY + pyte；缺 python/pyte 时跳过）。结果落 bench/results/，不入库 |
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.
- 12d ago Changed · +5 tokens per session 4cd95caf96cf
- 13d ago Changed · +8 lines · +262 tokens per session 54d049088d1c
- 14d ago Changed · +18 lines · +911 tokens per session scan A → C 8897c515dcee
- 15d ago Changed · +3 lines · +120 tokens per session fa3461cff852
- 23d ago Changed · +80 lines · +2,436 tokens per session 5bee2121bce5
- 24d ago Changed · +23 lines · +467 tokens per session 2416e2c4833a
- 25d ago Changed · +4 lines · +143 tokens per session 0a90006ccf4e
- 1mo ago Changed · -103 lines · -3,170 tokens per session fce28a2d3a96
- 1mo ago First seen · 172 lines · 4,297 tokens per session scan A c7b49b863fa7
moss AGENTS.md is an instructions file published in the GitHub repository D-Robotics/moss (143 stars, last pushed 5d ago), licensed MIT. It adds 5,471 tokens to every session, about $0.0219 per session on Opus 5.5. A static security scan graded it C with 2 findings (reaches for credential files, runs shell commands). No closer match exists in the catalogue, so it is treated as the original; first seen 2026-08-30.
Other instructions, from other repositories
vscode buildNext.instructions.md
Working notes and architecture documentation for the new esbuild-based build system in build/next. Use when making changes to the new build pipeline (transpile/bundle commands, NLS plugin, source-map handling, resource copying, or self-hosting watch tasks).
next.js AGENTS.md
AGENTS.md instructions for vercel/next.js, covering next.js development guide, codebase structure, monorepo overview, core package: packages/next and other important packages.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, cache eligibility and rollout and applying the notice (cutover).
langchain AGENTS.md
AGENTS.md instructions for langchain-ai/langchain, covering global development guidelines for the langchain monorepo, corridor security analysis, project architecture and context, monorepo structure and development tools & commands.
deepseek-harness AGENTS.md
AGENTS.md instructions for deepseek-ai/deepseek-harness, covering agents.md, pre-stable apis and released session data, repository layout, commands and host sandbox failures.
vscode copilot-instructions.md
Copilot instructions for microsoft/vscode, covering vs code copilot instructions, project overview, root folders, core architecture (src/ folder) and built-in extensions (extensions/ folder).