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/vcnoc/claude-code-zen-mcp-skill-work/agents-mdgit clone --depth 1 https://github.com/VCnoC/Claude-Code-Zen-mcp-Skill-WorkWrote 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/vcnoc/claude-code-zen-mcp-skill-work/agents-md)<a href="https://agentmods.dev/instructions/vcnoc/claude-code-zen-mcp-skill-work/agents-md"><img src="https://agentmods.dev/badge/instructions/vcnoc/claude-code-zen-mcp-skill-work/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 | $0.15836 | $0.15836 |
| Opus 5 | $0.07918 | $0.07918 |
| Sonnet 5 | $0.03167 | $0.03167 |
| Haiku 4.5 | $0.01584 | $0.01584 |
Grade A, and why
Claude-Code-Zen-mcp-Skill-Work 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 5d 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.
Copies of this mod
1 near-identical copy found in the catalogue:
- Claude-Code-Zen-mcp-Skill-Work CLAUDE.md — 100% identical, 0 lines differ
How it starts
The opening of the file, as written. The whole thing — 1,017 lines — stays where its author put it; the contents beside it link to each section on GitHub.
-
CLAUDE.md(Router → Phases)面向 AI 编程智能体的「轻量路由 + 多阶段 + 知识库驱动」规则集(项目配置文件)
目的:基于全局规则和路由机制处理当前用户消息,结合 P1-P3 阶段规则进行响应(P4 不参与初始路由),参考
项目知识库内容结构与生成规则统一模板生成正确的项目知识库文档。
📚 共享概念速查(SSOT)
以下概念在整个规则体系中只此一处定义,所有技能文档必须引用而非重复。
automation_mode (自动化模式)
核心理念:全自动模式 = 零等待原则(Zero-Wait Principle)
automation_mode=true 的本质含义:
- 用户说"全程自动化" = 用户已将所有决策权委托给 AI
- 除阻塞性错误(环境缺失、依赖错误、安全风险)外,禁止在任何节点等待用户确认
- "零等待"不是建议,是强制约束 - 违反此原则 = 违背用户授权意图
记忆锚点:
- automation_mode=true → 像执行 shell 脚本一样自动运行到底
- 技能返回 = 脚本执行到下一行,不是暂停点
- 询问用户 = 违反原则(除非阻塞性错误)
技术规范:
- 谁设: 仅 main-router 在任务开始时判断并设置
- 触发关键词: "全程自动化" / "自动化流程" / "全自动" / "自动化模式"
- 取值:
true(全自动) /false(交互,默认) - 传递:
[AUTOMATION_MODE: true](通过上下文) - 下游约束: 只读取,不重新判断;true 时禁止询问用户"是否继续?"
- 例外情况: 阻塞性错误(环境缺失、依赖错误、权限问题)、安全风险(敏感信息暴露、生产环境操作)
coverage_target (测试覆盖率目标)
- 谁设: 仅 main-router 在 P1/P2 阶段询问用户并设置
- 默认: 85% (工业标准) / 最低: 70%
- 传递:
[COVERAGE_TARGET: 85%](通过上下文) - 下游约束: 只读取,不询问用户;验证时使用该值作为标准
- 行为规范: 见 [G9 测试覆盖率目标设定] (包含 < 70% 强制报错等逻辑)
auto_log (自动化决策日志)
- 触发: 仅当 automation_mode=true 时生成
- 三层架构:
- Layer 1 核心模板 (约200行):
skills/shared/auto_log_template.md- 常驻内存,提供7个必选章节骨架 - Layer 2 详细规范 (约400行):
skills/shared/auto_log_detailed_spec.md- 按需引用,包含信息提取规则、质量检查清单、FAQ - Layer 3 示例库 (约300行):
skills/shared/auto_log_examples.md- 仅参考时查阅,包含完整示例和决策树可视化
- Layer 1 核心模板 (约200行):
- 职责分工: 技能输出片段 → router 汇总 → simple-gemini 生成 auto_log.md
- 文件性质: 运行时审计日志,不纳入版本控制
- Layer选择标准(自动判断):
- 仅Layer 1:简单任务(耗时<30分钟 且 阶段≤3 且 无P4触发)
- Layer 1+2:复杂任务(耗时≥30分钟 或 含P4阶段 或 质量问题≥5个)
- Layer 1+2+3:需要参考示例时(首次生成 或 复杂决策树可视化需求)
复杂操作规范(独立文档)
以下规范因操作步骤复杂,抽取为独立文件:
- G10 环境自适应 CLI 调用: 见
references/standards/cli_env_g10.md - P4 最终质量验证: 见
references/standards/p4_final_validation.md
核心工作流(优先执行)
智能技能路由优先原则:
在处理任何用户请求之前,应优先使用 main-router skill 进行智能路由和技能匹配。main-router 将基于以下标准自动选择最合适的工具或技能:
- 标准文件:全局和项目级的 CLAUDE.md
- 当前阶段:P1 (分析问题) / P2 (制定方案) / P3 (执行方案) / P4 (错误处理)
- 用户意图:问答、深度分析、代码审查、文档生成、规划制定等
可用技能/工具:
zen-chat- 一般问答和概念解释zen-thinkdeep- 复杂问题深度调查codex-code-reviewer- 代码质量审查(5 维度检查)[代码完成后强制使用]simple-gemini- 标准文档和测试代码生成 [文档/测试生成强制使用]deep-gemini- 深度技术分析文档(含复杂度分析)plan-down- 智能规划与任务分解 [plan.md 生成强制使用]
职责分配(Responsibility Matrix):
角色 职责 执行时机 调用方式 主模型(Claude Code 主会话) - 接收用户请求- 调用 main-router skill(任务开始时)- 调用其他技能(根据需要)- 执行自动恢复检查(技能返回后)- 创建和更新 TodoList- 输出结果给用户 整个任务生命周期 - main-router skill - 意图识别- 阶段匹配- 设置 automation_mode- 选择最佳技能 任务开始时(一次性) 主模型调用 → 返回建议 plan-down skill - 方法清晰度判断- 任务分解- 生成 plan.md 需要规划时(一次性) 主模型调用 → 返回 plan.md codex skill - 代码质量检查- 5 维度验证 代码完成后(一次性或多次) 主模型调用 → 返回检查报告 simple-gemini skill - 文档生成- 测试生成 需要文档/测试时(多次) 主模型调用 → 返回文档/测试代码 关键点:
- 所有 skill 都是一次性调用,执行完返回给主模型
- 主模型负责整个流程的控制和自动恢复
- main-router 负责初始路由和全程监控,在关键节点主动调用专用技能(遵循 G11 Anti-Lazy 原则)
职责分工(符合 G11):
- main-router:初始路由 + 全程监控 + 强制调用技能(Anti-Lazy 原则)
- 主模型:执行任务 + 技能返回后自动恢复 + 遵循 router 监控指令
- 技能(skills):一次性调用,返回给主模型,不持续运行
工作流程(含自动化模式状态管理):
用户请求 ↓ 主模型调用 main-router skill ↓ main-router 读取标准文件 (CLAUDE.md) ↓ CRITICAL:判断并设置 automation_mode 状态标志 - 检测关键词:"全程自动化" / "自动化流程" / "全自动" / "自动化模式" - 设置全局状态: automation_mode = true/false - 该状态在整个任务生命周期中保持不变 ↓ 意图分析 + 阶段匹配 + 置信度评分 ↓ 选择最佳技能/工具 (或直接执行) ↓ main-router 返回建议给主模型 ↓ 主模型执行任务 ← main-router 持续监控(Anti-Lazy),主模型负责自动恢复 - 所有下游技能从上下文中读取 automation_mode - 禁止下游技能重新判断自动化模式 - 技能返回后立即执行自动恢复检查 ↓ 关键节点 main-router 强制调用技能(主模型执行): - 代码完成 → 调用 codex skill(继承 automation_mode) - 需要测试 → 调用 simple-gemini skill → codex 验证(继承 automation_mode) - 需要规划 → 调用 plan-down skill(继承 automation_mode) - 需要文档 → 调用 simple-gemini/deep-gemini skill(继承 automation_mode)
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.
- 5d ago First seen · 1,017 lines · 15,836 tokens per session scan A 0a99748dd7cf
Claude-Code-Zen-mcp-Skill-Work AGENTS.md is an instructions file published in the GitHub repository VCnoC/Claude-Code-Zen-mcp-Skill-Work (116 stars, last pushed 8mo ago), licensed Apache-2.0. It adds 15,836 tokens to every session, about $0.0792 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
codex AGENTS.md
AGENTS.md instructions for openai/codex, covering rust/codex-rs, the codex-core crate, code review rules, crate api surface and model visible context.
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).
spec-kit AGENTS.md
AGENTS.md instructions for github/spec-kit, covering agents.md, about spec kit and specify, quickstart — add a new integration in 5 steps, integration architecture and integrationmanifest — file tracking.
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.
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.
vscode oss-third-party-notices.instructions.md
Instructions for microsoft/vscode, covering vs code oss third-party-notices pipeline, architecture, pipeline flow in ci, applying the notice (cutover) and fallback chain (never fail the build).