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 skills add yugef3h/leo-skills --skill sop-designgit clone --depth 1 https://github.com/yugef3h/leo-skillsWrote 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/skills/yugef3h/leo-skills/sop-design)<a href="https://agentmods.dev/skills/yugef3h/leo-skills/sop-design"><img src="https://agentmods.dev/badge/skills/yugef3h/leo-skills/sop-design/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/skills/yugef3h/leo-skills/sop-design"><img src="https://agentmods.dev/badge/skills/yugef3h/leo-skills/sop-design.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 | $0.00093 | $0.02338 |
| Opus 5 | $0.00046 | $0.01169 |
| Sonnet 5 | $0.00019 | $0.00468 |
| Haiku 4.5 | $0.00009 | $0.00234 |
Grade A, and why
sop-design 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 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.
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 — 166 lines — stays where its author put it; the contents beside it link to each section on GitHub.
SOP Design — Spec + Harness 双支柱开发流程
Overview
将 AI 辅助开发工程化为两个支柱:Spec(文档先行,定义"做什么") 和 Harness(质量保障,验证"做对了没")。核心理念:让 AI 自己写 spec、自己拆 task、自己写代码,Harness 自动化验收——人只做 review 和方向引导。
技术栈: React 前端 + Python/Go 后端
🚨 五大风险警告
以下风险必须在每个生成的 spec 文档中引用并评估等级。忽视任何一项都将导致项目失控。
| # | 风险 | 根因 | 应对 |
|---|---|---|---|
| 🔴 1 | AGENTS.md 过大 = 无规则 | 规则越多,AI 越容易跑偏,等同于没有规则 | 三次原则写入 + 大小监控(>200 行必须压缩)+ 按需加载。见 references/claude-setup.md |
| 🔴 2 | 先代码后文档 = 浪费时间 | 直接让 AI 写代码,功能可能对但架构全错,无法维护 | Spec→Plan→Task→Code 绝对顺序,写入 AGENTS 强制执行。违反时 Harness 拒绝通过 |
| 🟡 3 | AI 写测试需分层对待 | AI 生成的单元/集成测试已足够可靠,但 E2E 测试仍不稳定 | 分层策略:静态分析 + 单测默认开启,集成测试关键行为开启,E2E 默认关闭。见 references/testing-trophy.md |
| 🟡 4 | 巨型 Task → 难以 review | Task 太粗,一次生成太多代码,review 困难,改错成本高 | 每个 Task 应在 50-150 行代码量,一个 Task 一次 Conventional Commit |
| 🟡 5 | 功能都对/架构全错 | AI 生成的代码能跑,但不符合架构设计,维护成本飙升 | Harness 审查必须核对 spec 一致性(不只是功能正确);定期触发架构重构 |
核心铁律(写入 AGENTS.md)
1. 文档先行:Spec → Plan → Task → Code,不可逆。任何代码生成前必须先有对应 spec。
2. 三次原则:同一问题/模式出现 ≥3 次,才写入 AGENTS.md。杜绝一次教训就上规则。
3. 测试分层:静态分析始终开启,单元测试默认开启,集成测试关键行为开启,E2E 默认关闭。
4. 即改即提:一个 Task 完成 → 一个 Conventional Commit。不要把对的改错了。
5. 不写自然语言:生成代码时只输出代码,不在代码中夹杂解释性自然语言。
模块路由
先判断用户意图,再加载对应 reference:
| 用户说什么 | 激活模块 | 加载文件 |
|---|---|---|
| "新项目""没有 CLAUDE.md""/init""初始化" | 项目初始化 | references/claude-setup.md |
| "新功能""写个 spec""需求文档""技术方案" | Spec 支柱 | references/spec-pillar.md + references/bdd-gherkin.md |
| "审查""测试""重构""AGENTS 太大了""加个规则" | Harness 支柱 | references/harness-pillar.md + references/testing-trophy.md |
| "提交""commit""task 拆完了" | 交付节奏 | references/conventional-commits.md |
| 不确定 | 先问:要写 spec 定义"做什么",还是用 Harness 验证"做对了没"? | — |
双支柱总览
用户需求
│
├── Spec 支柱(文档先行)──────────────────────┐
│ Requirements Spec ──▶ Design Spec │
│ (BDD/Gherkin) (C4/架构图) │
│ │ │ │
│ └──────┬──────────────┘ │
│ ▼ │
│ Implementation Spec │
│ (Task 拆解 + 依赖 + Checklist) │
│ │ │
│ ▼ │
│ AI 生成代码 │
│ │ │
└───────────────┼───────────────────────────────┘
▼
┌── Harness 支柱(质量保障)────────────────────┐
│ │
│ TDD ──▶ Code Review ──▶ Refactor │
│ (分层) (Spec一致性) (人工引导) │
│ │
│ AGENTS.md 维护(三次原则 + 压缩) │
│ │
│ 反馈 ──▶ 更新 Spec(形成闭环) │
└───────────────────────────────────────────────┘
What ships with it
17 files beside SKILL.md in the same directory: the scripts, references and assets a skill reads on demand. Not counted in the per-session cost; read them before you install if any of them is executable.
- CLAUDE_README.md 3.6 KB
- claude-config.md 6.3 KB
- other-important.md 147 B
- references/bdd-gherkin.md 5.6 KB
- references/claude-setup.md 4.3 KB
- references/conventional-commits.md 4.6 KB
- references/harness-pillar.md 11 KB
- references/spec-pillar.md 10 KB
- references/testing-trophy.md 8.6 KB
- static/Harness.png 613 KB
- static/README.md 2.3 KB
- static/Spec.png 425 KB
- static/TDD.png 291 KB
- TODO.md 6.8 KB
- vibeCoding高质量作品完整方案.md 7.6 KB
- 前端工程化约束.md 6.1 KB
- 组件思考.md 7.0 KB
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 First seen · 166 lines · 93 tokens per session scan A f24866a8287e
sop-design is a skill published in the GitHub repository yugef3h/leo-skills (11 stars, last pushed 5d ago), licensed MIT. It adds 93 tokens to every session and 2,338 once invoked, about $0.0005 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 skills, from other repositories
agent-team
An AI team-building skill for complex projects, with roles for execution, direction, and review. It lets you follow discussions and decisions while the team shares the same project context.
chrome-automation
A skill for automating an already open Google Chrome browser through its browser control connection.
digital-avatar-shopping-video
A digital-avatar video maker for product recommendations and promotional sales videos. It creates a spoken presentation by coordinating the script, visuals, and audio-video assembly.
obsidian-bases
An Obsidian skill for creating database-view files. Obsidian is a note-taking app, and a database view shows notes as tables or cards using filters and formulas.
product-marketing-copywriter
A marketing-copy tool that analyzes audience problems and product benefits, then creates promotional headlines and body text. Marketing copy is writing intended to explain and promote a product.
atutun-xhs-cover
A prompt generator for creating Xiaohongshu cover designs, where Xiaohongshu is a Chinese social platform for lifestyle and product content.