From research-survey
Build rigorous, problem-driven research surveys for a specific research direction. Use when the user asks to survey a field, map related work, identify concrete research problems, analyze technical challenges, compare innovations, find gaps, design 'what we should do next', write survey notes, build a literature review, or connect a direction to their own research. Supports effort levels (lite/standard/deep) and anti-hallucination verification.
How this skill is triggered — by the user, by Claude, or both
Slash command
/research-survey:research-surveyThe summary Claude sees in its skill listing — used to decide when to auto-load this skill
Turn a research direction into a rigorous survey organized around concrete problems, challenges, prior solutions, remaining gaps, and possible next research moves. Prioritize a field-level logic chain over a paper-by-paper summary.
Turn a research direction into a rigorous survey organized around concrete problems, challenges, prior solutions, remaining gaps, and possible next research moves. Prioritize a field-level logic chain over a paper-by-paper summary.
Write the final survey in Chinese by default unless the user asks for another language.
Do not start from "what papers exist"; start from "what concrete problem must this specific direction solve." The direction must be narrow enough to make claims precise.
Good scope: physics-informed neural operators for long-horizon spatiotemporal forecasting with limited sensors
Bad scope: deep learning, PDE, control
Choose effort level from the user's wording or explicitly stated preference:
| Level | Trigger | Scope | Paper Count | Depth |
|---|---|---|---|---|
| lite | "简单调研", "quick survey", "大概看看" | Single problem, 1-2 method families | 5-10 papers | Challenge map + gap identification |
| standard | Default, "做个调研", "survey this" | Full problem chain, all major method families | 15-30 papers | Full template with method comparison |
| deep | "深度调研", "exhaustive survey", "写综述" | Comprehensive coverage, historical evolution | 30-60+ papers | Full template + timeline + open problems + detailed method cards |
Search Strategy (ordered by reliability):
/arxiv or user input/semantic-scholar/openalex and /research-lit/deepxiv for layered access (search→abstract→section)/arxiv explicitly for papers from the last 12 monthsLiterature Search Skills (from ARIS, use as needed):
| Skill | When to use |
|---|---|
/arxiv | Preprint search, PDF download, latest unrefereed work |
/semantic-scholar | Published venue papers (IEEE, ACM, Springer), citation counts |
/openalex | Broad open academic metadata, citation graphs |
/deepxiv | Progressive section-level reading without loading full papers |
/alphaxiv | Quick LLM-optimized summary of a single paper |
/research-lit | Multi-source aggregated search (combines multiple sources) |
/exa-search | AI web search beyond academic DBs (requires Exa key) |
Search execution order for standard effort:
/arxiv "topic" → get recent preprints (top 10-15)/semantic-scholar "topic" → get published venue papers (top 10)/deepxiv for progressive reading/openalex "topic" → fill gaps in coverageFor each paper, record:
Anti-Hallucination Rules:
[未验证]paper-digest skill only for papers you have actually accessed (PDF, abstract, or full text)Before writing any output, build the logic chain internally:
Field → Concrete Problem → Why Hard (Challenges) →
Prior Solutions (Method Families) → What Each Solves →
What Remains (Gaps) → Possible Next Work
Verification checklist:
Choose the survey lens:
Before delivering the final survey:
Use when the survey should support research positioning, proposal writing, or "how should we do our work?"
这个具体领域要解决什么问题
Related work 如何解决这些问题
Related work 还有什么不足
我们要怎么做来解决这些不足
Use when the survey should expose the technical skeleton and map innovations to challenges.
这个具体领域要解决什么问题
解决问题有什么挑战
相关工作提出什么创新点解决什么挑战
现在还有什么问题
paper-digest for deep notes on key papers (seminal, SOTA, complex methods)research-context for user's research direction (for inspiration/transfer sections)Use for standard and deep effort levels. Compress for lite level.
# [MMDD]·[Survey] 具体研究方向
> 核心摘要 (One-Liner):
该方向试图解决 [具体领域/任务] 中的 [具体问题];现有工作主要通过 [主要方法族] 解决 [关键挑战],但仍存在 [核心不足]。潜在突破口是 [方向]。
| 项目 | 内容 |
| --- | --- |
| **具体方向** | [狭义研究方向] |
| **核心问题** | [该方向最核心、最具体的问题] |
| **关键任务/场景** | [任务、数据、输入输出、约束] |
| **主要挑战** | [Challenge 1; Challenge 2; ...] |
| **代表论文/系统** | [代表作,含年份] |
| **可用代码/基准** | [代码、benchmark、dataset] |
| **调研深度** | lite / standard / deep |
| **论文覆盖** | [X 篇已验证, Y 篇未验证] |
---
## 1. 领域定义与核心问题
- **领域边界:** [为什么这个方向足够具体?]
- **核心问题:** [一句话定义]
- **问题的重要性:** [技术/工程/应用意义]
- **现有范式的不足:** [为什么传统方法不够?]
---
## 2. 关键挑战分解
| 挑战 | 为什么难 | 现有方法为何不足 | 需要的能力 |
| --- | --- | --- | --- |
| Challenge 1 | [原因] | [不足] | [能力] |
| Challenge 2 | [原因] | [不足] | [能力] |
---
## 3. Related Work 如何解决问题
### 3.1 方法族 A:[名称]
- **解决的挑战:** [Challenge X]
- **核心创新:** [设计决策]
- **技术机制:** [模型/算法/训练/数据/公式]
- **代表工作:** [论文 1 (venue'YY), 论文 2 (venue'YY)]
- **证据:** [实验支持了什么]
- **不足:** [该方法族仍未解决什么]
### 3.2 方法族 B:[名称]
[同上结构]
---
## 4. 挑战-创新-证据映射表
| 挑战 | 代表工作 | 创新点 | 解决机制 | 实验证据 | 剩余不足 |
| --- | --- | --- | --- | --- | --- |
| [Challenge] | [Paper] | [Innovation] | [Mechanism] | [Evidence] | [Gap] |
---
## 5. 方法对比分析 (standard/deep only)
| 维度 | 方法族 A | 方法族 B | 方法族 C |
| --- | --- | --- | --- |
| 核心思路 | | | |
| 解决的挑战 | | | |
| 计算成本 | | | |
| 数据需求 | | | |
| 泛化能力 | | | |
| 主要局限 | | | |
---
## 6. 现在还有什么问题
- **Field-level Gap 1:** [跨论文共同不足]
- **Field-level Gap 2:** [跨方法族的未解挑战]
- **Evaluation Gap:** [缺失数据/指标/场景]
- **Theory/Mechanism Gap:** [解释/保证/可控性不足]
- **矛盾与 Trade-off:** [方法 A 改善 X 但恶化 Y]
---
## 7. 我们要怎么做
- **研究切入点:** [从哪个未解问题切入]
- **核心想法:** [设计决策 + 为什么可能有效]
- **对应挑战:** [解决哪个 Challenge]
- **技术路线草案:** [输入→处理→输出→训练→评估]
- **最小可行实验:** [dataset/benchmark/baseline/metric]
- **预期贡献:** [问题/方法/实验/理论贡献]
- **风险与验证:** [可能失败处;什么结果才能证明有效]
- **与我的研究方向的关联:** [参考 research-context]
---
## 8. 论文清单与阅读优先级
| 优先级 | 论文 | Venue/Year | 为什么要读 | 用 paper-digest 深读 | 验证状态 |
| --- | --- | --- | --- | --- | --- |
| P0 | [核心/奠基/SOTA] | [venue'YY] | [原因] | 是/否 | ✓/[未验证] |
| P1 | [补充] | [venue'YY] | [原因] | 是/否 | ✓/[未验证] |
---
## 9. 总结
- **一句话判断:** [该方向是否值得做,为什么]
- **最值得突破的问题:** [具体问题]
- **下一步行动:** [搜索/精读/实验/复现/写proposal]
Before finalizing, verify:
Provides UI/UX resources: 50+ styles, color palettes, font pairings, guidelines, charts for web/mobile across React, Next.js, Vue, Svelte, Tailwind, React Native, Flutter. Aids planning, building, reviewing interfaces.
Fetches up-to-date documentation from Context7 for libraries and frameworks like React, Next.js, Prisma. Use for setup questions, API references, and code examples.
npx claudepluginhub huaer02/research-skills --plugin research-survey