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 vanbuster/vanceling-skills --skill ai-paper-readinggit clone --depth 1 https://github.com/vanbuster/vanceling-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/vanbuster/vanceling-skills/ai-paper-reading)<a href="https://agentmods.dev/skills/vanbuster/vanceling-skills/ai-paper-reading"><img src="https://agentmods.dev/badge/skills/vanbuster/vanceling-skills/ai-paper-reading/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/vanbuster/vanceling-skills/ai-paper-reading"><img src="https://agentmods.dev/badge/skills/vanbuster/vanceling-skills/ai-paper-reading.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 | $0.00097 | $0.02054 |
| Opus 5 | $0.00048 | $0.01027 |
| Sonnet 5 | $0.00019 | $0.00411 |
| Haiku 4.5 | $0.00010 | $0.00205 |
Grade A, and why
ai-paper-reading 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 11d 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 — 218 lines — stays where its author put it; the contents beside it link to each section on GitHub.
AI 论文精读 Skill
适用场景
- 需要深度理解一篇 AI/LLM/Agent/RAG 相关论文
- 准备面试，需要能"用人话"讲清楚某篇论文的核心思想
- 快速跟进一个新领域，建立技术理解框架
触发词
"读论文"、"帮我拆解这篇论文"、"这篇论文讲了什么"、"精读"
四阶段流程（30 分钟 / 篇）
选论文 → AI 辅助预读 → 结构化拆解 → 费曼输出 → 边界检测
2min 5min 12min 8min 3min
阶段 0：选论文（2 min）
原则
- 首选源头论文（如 Transformer 看 "Attention Is All You Need"），不先从综述/解读入手
- 被引量 > 100 的论文优先
- 有开源代码的论文优先（代码 + 论文对照读）
不要做的
- 不要从 medium 博客、知乎文章开始（那些是二手信息）
- 不要试图逐字读完整篇论文（大部分内容 AI 可以帮你提取）
阶段 1：AI 辅助预读（5 min）
操作步骤
- 把论文 PDF 喂给 AI（Claude/ChatGPT/DeepSeek）
- 使用以下 prompt 快速建立全局认知：
请用 5 句话以内告诉我：
1. 这篇论文要解决什么问题？
2. 核心方法是什么（一句话，不用公式）？
3. 跟之前的方法比，关键差异是什么？
4. 实验效果提升了多少（给一个数字）？
5. 最大的局限性或未解决的问题是什么？
预读检查点
- 能用一句话说出这篇论文在做什么
- 知道它的方法叫什么名字
- 知道它比 baseline 提升的大致幅度
如果预读后 3 个问题都答不上来 → 换一篇，这篇可能不适合你当前水平
阶段 2：结构化拆解（12 min）
使用以下 Prompt（直接复制给 AI）
请以"严格的论文审稿人"身份，精读这篇 AI 论文。
你的目标是帮我彻底理解它的技术逻辑。
## Q1 问题定位
它要解决什么痛点？这个痛点是真实存在的还是假设的？
（判断标准：论文是否给出了具体的失败案例或数据来证明痛点存在？）
## Q2 方法核心
解决方案的核心思路是什么？
输入是什么、经过什么处理、输出是什么？
不要抛公式，用自然语言描述数据流转。
## Q3 关键差异
它和之前的主流方法有什么本质不同？
（追问：如果只是换了个更大的模型/更多的数据，不算本质不同）
## Q4 实验证据
用什么数据集？评估指标是什么？结果比 baseline 好多少？
（追问：这个提升幅度在工程实践中是否有意义，还是纯学术刷点？）
## Q5 祛魅与局限
这篇论文的局限性或潜在问题是什么？
什么情况下这个方法会失效？
（要求：至少列出 2 个局限，不能只说"未来可以改进"）
## Q6 知识关联
它和以下哪个概念有直接关系？是什么关系？
- RAG（检索增强生成）
- Agent（感知/规划/记忆/工具调用）
- Transformer / 注意力机制
- Prompt 工程
- RLHF / 对齐
- MoE（混合专家）
（如果都不相关，请说明它属于哪个独立方向）
## Q7 直觉类比
用日常生活中的类比（做饭、开车、搭积木、图书馆找书等）
来解释这篇论文的核心思想。要让完全不懂 AI 的人也能听懂。
## Q8 检验题
设计 2 道题检验我是否真懂了：
- 概念题：检验我是否理解了核心方法
- 反事实题："如果去掉 X 模块，结果会怎样？"
检验我是否理解了设计选择的理由
## 约束
- 不要概括，要拆解
- 不要写"本文提出了一个创新方法"这类空话
- 术语保留英文原文（如 "Retrieval-Augmented Generation"）
- Q8 的答案和你对答案的解释必须分开放
- 每个回答控制在 150 字以内
拆解检查点
- Q1-Q8 全部有明确答案
- Q7 的类比你能用自己的话复述
- Q8 的两道题你能答对
阶段 3：费曼输出（8 min）
操作步骤
- 用嘴说出来（不是默读、不是打字），向一个"假想的同事"解释这篇论文
- 如果卡壳 → 标记那个点，回去看 AI 的拆解
- 如果解释得太复杂 → 问 AI "请用更简单的类比再解释一次 X"
- 录一段 2 分钟语音，自己听一遍
费曼检查 Prompt
我刚才对这篇论文的理解是这样的：
[粘贴你的理解]
请指出：
1. 我哪里说错了？
2. 我遗漏了什么关键点？
3. 有没有更好的类比可以替换我用的那个？
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.
- 11d ago First seen · 218 lines · 97 tokens per session scan A 94e12f23b3fe
ai-paper-reading is a skill published in the GitHub repository vanbuster/vanceling-skills (9 stars, last pushed 2mo ago), licensed MIT. It adds 97 tokens to every session and 2,054 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 2026-08-31.
Other skills, from other repositories
understand-explain
Use when you need a deep-dive explanation of a specific file, function, or module in the codebase.
baoyu-comic
A tool for creating original educational comics that explain knowledge or ideas through multiple illustrated panels. It supports different art styles and tones and can create several comics in one batch.
skill-creation-walkthrough
Step-by-step guide for creating your own Claude Skills, from deciding whether a skill is the right tool to writing the SKILL.md file, structuring reference material, and making it trigger reliably. Use when you want to package a workflow, framework, or repeated task into a reusable Skill, when an existing skill is not…
elixir-idioms
OTP/BEAM patterns and Elixir idioms — GenServer, Supervisor, Task, Registry, pattern matching, with chains, pipes. Use when designing processes or debugging BEAM issues.
master-atisha
A reference-based assistant for Atisha and the Kadam school, an early Tibetan Buddhist tradition. It explains the staged path to awakening and can answer in Atisha’s historical teaching style.
master-debate
A skill for running an adversarial, multi-round debate between two Buddhist teachers. It is designed for questions where the teachers argue from different positions.