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 yunshu0909/yunshu_skillshub --skill system-studygit clone --depth 1 https://github.com/yunshu0909/yunshu_skillshubWrote 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/yunshu0909/yunshu_skillshub/system-study)<a href="https://agentmods.dev/skills/yunshu0909/yunshu_skillshub/system-study"><img src="https://agentmods.dev/badge/skills/yunshu0909/yunshu_skillshub/system-study/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/yunshu0909/yunshu_skillshub/system-study"><img src="https://agentmods.dev/badge/skills/yunshu0909/yunshu_skillshub/system-study.svg" alt="Reviewed on agentmods" width="80" height="20"></a>- NVIDIA SkillSpector pass
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 | $0.00186 | $0.03108 |
| Opus 5.5 | $0.00074 | $0.01243 |
| Sonnet 5.5 | $0.00037 | $0.00622 |
| Haiku 4.5 | $0.00019 | $0.00311 |
Grade A, and why
system-study 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 1mo 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 — 177 lines — stays where its author put it; the contents beside it link to each section on GitHub.
System Study — 系统化学习材料生成
工具适配
文中的 AskUserQuestion、Agent、Read、Write 是能力名称，使用当前平台的等效工具。没有多选界面时，用一条消息列出切片选项，请用户回复编号组合；已明确的信息不重复问。子 agent 按平台并发上限分批执行，先固定全部输入和 prompt，后批不得注入前批的结论。平台没有子 agent 能力时说明限制，再与用户确定逐视角调研方式。
你要学一个新领域。AI 自主调研 + 搭体系 + 产出 HTML 学习材料。
和已有 skill 的本质区别：
- 与结论型调研的区别：结论型调研是"有困惑 → 求结论"；这个是"想学 → 求体系"
- vs case-radar：那个给散点案例集；这个给系统化全貌
- vs readable-output：那个处理已有素材；这个自己挖素材 + 自己建体系（最后可调用它做出稿，但不是核心）
这个 skill 的差异化命门：除了"骨架 + 案例 + 工程化"这些教科书必讲的部分，强制产出"争议焦点 + 盲区清单"两块——这是普通学习材料没有的，是用户研究 → 内容/产品的金矿。
5 阶段流程总览
| 阶段 | 名称 | 介入次数 | 产物 | 详细操作 |
|---|---|---|---|---|
| 1 | 题目收敛 | 1 次（必） | 01-题目确认.md |
reference/stage-1-topic-narrow.md |
| 2 | 调研计划 | 1 次（默认） | 02-调研计划.md |
reference/stage-2-research-plan.md |
| 3 | Sub-agent 并行调研 | 0 次 | 03-调研笔记/A-E.md |
reference/stage-3-subagent-templates.md ⭐ |
| 4 | 体系搭建+大纲确认 | 1 次（默认） | 04-体系大纲.md |
reference/stage-4-synthesis.md |
| 5 | HTML 学习材料 | 0 次 | 05-学习材料.html |
reference/stage-5-html-spec.md |
默认严格逻辑：阶段 1/2/4 都问用户。但用户可一次性说"全程不打扰、直接出 HTML"，跳过 2、4。阶段 1 永远不能省——题目错了后面全白干。
用户授权"全程不打扰"快速通道
如果用户在阶段 1 / 2 / 任何阶段说出"你直接来吧 / 直接给我结果就行 / 你自己定 / 你判断 / 全程不打扰"——主线程立即切换到快速通道：
- 阶段 1 依然要问（用 AskUserQuestion 弹切片）——题目永远不能 AI 替用户拍
- 阶段 2 仍然写
02-调研计划.md存档，但不再问用户，直接进阶段 3 - 阶段 4 仍然写
04-体系大纲.md存档，但不再问用户，直接进阶段 5 - 在 PLAN.md 标注"用户已授权全程不打扰"——这是审计线索
切换条件：用户的话里明确包含"直接 / 不打扰 / 自己定 / 自己判断 / 你来"等授权词。模糊表达（如"看着办"）不算——继续按默认严格流程问。
核心设计原则（5 条）
这 5 条是从首发实战提炼的、贯穿整个 skill 的设计哲学。所有 stage 都遵循：
- 必经卡点不省——阶段 1 题目收敛永远不能省。其他卡点用户可授权跳过，但题目错了全白干。
- 强制要异见，禁止和稀泥——Agent E 不接受"两边都对"。每个议题必须给 X/10 倾向分。
- 盲区清单独立成块——还没共识的、还在打架的、AI 互相复读的，独立段落，不混在其他内容里。
- 用户研究视角主导——不走"基础→进阶→高阶"教材路线，按"我能拿来做什么"组织。允许带判断、带视角，不需要装客观。
- HTML 独立写，不调 readable-output——学习材料需要侧栏导航 + 章节折叠 + 争议高亮 + 盲区独立区，长文阅读结构不够用。
What ships with it
7 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.
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.
- 1mo ago Changed · +8 lines · +8 tokens per session f505b4fdf8a5
- 2mo ago First seen · 169 lines · 178 tokens per session scan A 466b647e01f2
system-study is a skill published in the GitHub repository yunshu0909/yunshu_skillshub (789 stars, last pushed 26d ago), licensed MIT. It adds 186 tokens to every session and 3,108 once invoked, about $0.0007 per session on Opus 5.5. 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.
system-study compared with skills from other repositories
| Mod | Type | Stars | Last push | Licence | Scan | Tokens | Works with |
|---|---|---|---|---|---|---|---|
| system-studyyunshu0909/yunshu_skillshub | Skill | 789 | 26d ago | MIT | A | 186 | Claude Code, Codex |
| hr-onboardingnexu-io/open-design | Skill | 106k | yesterday | Apache-2.0 | A | 62 | Claude Code, Codex |
| book-mirrorgarrytan/gbrain | Skill | 32k | yesterday | MIT | A | 138 | Claude Code |
| exam-forecastanthropics/claude-for-legal | Skill | 9.9k | 3d ago | Apache-2.0 | A | 73 | Claude Code |
| lean-formalizewanshuiyin/Auto-claude-code-research-in-sleep | Skill | 19k | 3d ago | MIT | A | 81 | Claude Code |
| eli5companion-inc/feynman | Skill | 10k | 6d ago | MIT | A | 63 | Claude Code, Codex |
| deck-course-modulenexu-io/html-anything | Skill | 9.4k | 24d ago | Apache-2.0 | A | 25 | Claude Code, Codex |