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/libinyam/dsh-experts/claude-mdgit clone --depth 1 https://github.com/libinyam/dsh-expertsWrote 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/libinyam/dsh-experts/claude-md)<a href="https://agentmods.dev/instructions/libinyam/dsh-experts/claude-md"><img src="https://agentmods.dev/badge/instructions/libinyam/dsh-experts/claude-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 | $0.01799 | $0.01799 |
| Opus 5 | $0.00899 | $0.00899 |
| Sonnet 5 | $0.00360 | $0.00360 |
| Haiku 4.5 | $0.00180 | $0.00180 |
Grade A, and why
dsh-experts CLAUDE.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 — 51 lines — stays where its author put it; the contents beside it link to each section on GitHub.
dsh-experts 宪法
本文件是这个项目的最高工程约定。每次会话动手前先读本文件；宪法与代码冲突时，要么改代码，要么显式修订宪法并说明理由——不许静默偏离。
业务红线（违反即回滚，没有例外）
- 不猜 harness API。与 DeepSeek Harness 的每一个接点（服务名、方法签名、能力 flag、rank 语义）必须有 deepseek-harness 源码出处。本文件末尾的「接点台账」记录已验证接点及其出处；新接点先验证、记台账、再使用。查不到出处的能力：降级到已验证路径，或阻塞并告知用户，绝不臆测字段名。
- Fail loud，不静默降级。任何无效输入（坏 team.json、丢失卡片、非法路径）必须抛出带
code的错误，消息包含精确文件路径和原因。禁止：吞错后跳过、把无效团队静默从目录里消失、把「未采集」记成零分。用户修一个坏团队时，报错必须能一次定位到字段。 - 团队是数据，不是代码。引擎永不执行团队目录里的任何可执行内容；团队配置只允许结构化字段（见 manifest.js 的 schema）。不向团队开放自定义脚本/工作流代码——这条不因功能需求而放松，除非单独立项修订。
- 引擎只读。src/ 下的发现、校验、组装逻辑对文件系统只做读操作；唯一例外是 scripts/ 下用户显式执行的脚手架 CLI。引擎不做网络访问、不写任何文件。
- 路径即边界。团队清单里所有文件引用（card 等）必须在团队目录内解析。逃逸（
..、绝对路径、符号链接出目录）一律拒绝。违反它的测试必须存在且保持红敏感。
技术硬约定
| 约定 | 违反时什么会红 |
|---|---|
| 零运行时依赖、零 devDependencies；只用 Node 标准库 | guard: package.json 的 dependencies/devDependencies 非空即 fail |
| Plain ESM JavaScript，无构建步骤，无 TypeScript | guard: src/ 出现 .ts 文件或 tsconfig.json 即 fail；import 语法错误由 import 冒烟测试红 |
| Node >= 18 可运行（开发机 v24） | engines 字段 + CI 的 node 矩阵 |
所有抛错走 errors.js 的 pluginError(msg, code, details)，code 为稳定大写蛇值 |
代码评审清单 + guard: src/ 内 throw new Error( 直接出现即 fail |
错误码词汇表是 errors.js 导出的 ERROR_CODES（单一事实源） |
guard: 双向同步扫描——用了未声明的码、声明了未用的码都红 |
| manifest 校验拒绝未知字段（拼写保护）、拒绝路径逃逸（含 Windows 盘符前缀与 NUL）、强制恰好一个 coordinator；内联进技能 body 的自由文本字段必须单行且不含管道符；卡片不得含 4+ 反引号围栏行 | tests/manifest.test.js 对应反例必须红 |
| 发现层级 rank：project(100) < custom(300) < user(400) < bundled(600)，同名团队低 rank 覆盖（先读名去重，再对胜者全量校验——有效低 rank 团队可遮蔽同名坏团队） | tests/teams.test.js 层级覆盖与遮蔽用例 |
| 组装出的技能 body 必须含：花名册表、每专家人设卡围栏块、升级矩阵表、工作流模板、团队约定五节 | tests/compose.test.js 断言各节存在 |
| 专家发现协议只用保守 JSON Schema 子集（type/properties/items/required/additionalProperties/enum），作为 prompt 文本协议传给 specialist | 模板内 schema 由 tests/compose.test.js 解析断言子集 |
| src/ 禁 console.log（CLI 脚本除外） | guard: grep src/ 中 console. 即 fail |
| 文本文件 LF、UTF-8、两空格缩进 | .editorconfig + .gitattributes；guard 冒烟检查 BOM |
产品路线（已拍板）
- v0.1（本版）：lead-first review 工作流模板 + web-review 示例团队（1 个 lead + 4 个按需 specialist）+ 团队校验 CLI + 脚手架 CLI + bundle 注册。
- 明确不做（写入已知问题，防止顺手做出）：develop 工作流模板、文件 watcher（新增团队需重载插件）、
.agents系根目录、按专家强制模型路由（v0.1 仅模板提示）、团队市场/远程安装。 - 专家团运行时依赖 dsh 原生
subagent、list_agents、send_message和 child-scopedreport；lead 是真实 child，specialist 由 lead 按需启动。部署未装载这些能力时必须标记团队运行时不可用，禁止伪装成专家已执行。
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 · 51 lines · 1,799 tokens per session scan A 3f9f900fc576
dsh-experts CLAUDE.md is an instructions file published in the GitHub repository libinyam/dsh-experts (2 stars, last pushed 18d ago), licensed MIT. It adds 1,799 tokens to every session, about $0.0090 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 2026-08-31.
Other instructions, from other repositories
working-activity AGENTS.md
AGENTS.md instructions for ccch1mneyyy/working-activity, covering working-activity 项目规则, 用途, pi 版开发约定, dsh 版开发约定 and 文案.
dsh-codex-sync AGENTS.md
Instructions for Walvez/dsh-codex-sync, covering agents.md — dsh-codex-sync, 定位, 怎么跑, 技术栈 and 目录与约定.
causal-memory CLAUDE.md
Instructions for JingxuanC/causal-memory, covering causal memory mcp integration, causal memory integration, after acting on a decision and observing the result, when something fails unexpectedly and when a recorded lesson turns out to be wrong.
dsh-deepread AGENTS.md
Instructions for xiehuan123/dsh-deepread, covering agents.md, upstream authority, plugin identities and source files, release documentation and harness lifecycle rules.
dsh-checkpoint-rewind AGENTS.md
Instructions for PerryLink/dsh-checkpoint-rewind, covering agents.md, 仓库布局：发布面 / 本地工程面, 命令, 提交纪律 and dsh 插件约束（红线）.
dsh-industry-research AGENTS.md
Instructions for PerryLink/dsh-industry-research, covering agents.md, layout, hard rules applied here, checks and release.