in-depth-research-guide

本条目面向文献与学术工具场景,提供“in-depth-research-guide”技能的目录入口。公开标签:learning、多源调查、证据合成、学术检索。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

仓库
brycewang-stanford/Auto-Empirical-Research-Skills
仓库最近推送
Stars
5,023
Forks
578
Issues / PRs
0

安装方式

先用 Prompt 核对来源,也可切换到安装命令或下载可用的安装包。

安装方式

用 AI 助手安装

复制提示词,让助手先核对固定版本和依赖。

查看 Prompt收起 Prompt
先检查当前 Agent 是否已安装 SkillPI Skill Finder。如果尚未安装,请按官方仓库的说明安装并启用:https://github.com/Findskills/skillpi。然后使用 Skill Finder 打开此 SkillPI 详情页并定位 "in-depth-research-guide":https://skillpi.com/zh/skills/deepskill-09eebfba9af6c61c1964b2add8f556dc81fb9f36366d005f74abd2a1fe9c30aa。安装前核对上游来源和版本:https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/research/deep-research/in-depth-research-guide/SKILL.md。来源不一致时停止并说明差异。核对通过后,用 Skill Finder 安装目标 Skill,并报告安装路径和来源版本。

此命令从 GitHub 来源仓库获取指定技能。运行前请先查看源文件。

运行工具
npx skills add https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/tree/bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba/skills/43-wentorai-research-plugins/skills/research/deep-research/in-depth-research-guide --skill in-depth-research-guide
许可协议: —

来源与配置

查看 GitHub 源码

brycewang-stanford/Auto-Empirical-Research-Skills固定提交 bfa0a6e5a843babbf45a1ef1663fd5dc4f07c0ba

配置方式待核实

请查看原始来源中的适用工具、依赖和权限要求。

安装实测待补充

本条目面向文献与学术工具场景,提供“in-depth-research-guide”技能的目录入口。公开标签:learning、多源调查、证据合成、学术检索。请通过固定上游版本核对实际功能、依赖与权限;此页为 Skillπ 独立编写的用途介绍和原站链接,不代表运行安全保证。原文与附件依各自许可开放。

↓ 0 下载 ↗ 0 分享点击

文件浏览

在原站浏览文件 ↗

原始 SKILL.md

打开原文 ↗
---
name: in-depth-research-guide
description: "Structured methodology for conducting exhaustive multi-source investigations"
metadata:
  openclaw:
    emoji: "🔬"
    category: "research"
    subcategory: "deep-research"
    keywords: ["deep research", "systematic investigation", "multi-source research", "evidence synthesis", "research methodology", "source evaluation"]
    source: "wentor-research-plugins"
---

# In-Depth Research Methodology

## Overview

In-depth research goes beyond surface-level literature review to conduct exhaustive, multi-source investigations that synthesize evidence from academic papers, grey literature, industry reports, datasets, and primary sources. This methodology is used when a research question requires comprehensive coverage — for systematic reviews, policy briefs, competitive analyses, or foundational literature surveys in a new research direction.

## The 5-Phase Investigation Framework

### Phase 1: Scope Definition (10% of effort)

Before searching, define boundaries explicitly:

```markdown
## Research Brief Template

**Central Question**: [One sentence, specific and falsifiable]
**Sub-Questions** (3-5):
  1. [Decomposed aspect 1]
  2. [Decomposed aspect 2]
  3. [Decomposed aspect 3]

**Inclusion Criteria**:
  - Time range: [e.g., 2018-present]
  - Languages: [e.g., English, Chinese]
  - Document types: [peer-reviewed, preprints, reports, patents]
  - Disciplines: [e.g., CS, cognitive science, linguistics]

**Exclusion Criteria**:
  - [Opinion pieces, blog posts without data]
  - [Studies with n < 30 unless qualitative]
  - [Duplicate publications of same study]

**Expected Deliverable**: [Literature review / Evidence map / Policy brief / State-of-art report]
**Depth Target**: [Exhaustive / Representative / Exploratory]
```

### Phase 2: Multi-Source Collection (30% of effort)

Search systematically across source tiers:

| Tier | Source Type | Examples | Purpose |
|------|-----------|---------|---------|
| **1** | Academic databases | OpenAlex, PubMed, Scopus, Web of Science | Peer-reviewed primary research |
| **2** | Preprint servers | arXiv, bioRxiv, SSRN, medRxiv | Cutting-edge, not yet reviewed |
| **3** | Grey literature | WHO reports, World Bank, NBER working papers | Policy and institutional knowledge |
| **4** | Patents and standards | Google Patents, USPTO, IEEE standards | Technical implementations |
| **5** | Data repositories | Zenodo, Figshare, Kaggle, ICPSR | Raw data and reproducibility |
| **6** | Expert knowledge | Conference talks, interviews, personal communication | Tacit knowledge, emerging trends |

**Search strategy per source**:

```markdown
For each source:
1. Construct 3-5 query variants (synonyms, related terms, translated terms)
2. Apply inclusion/exclusion filters
3. Record: query string, date, results count, relevant hits
4. Download and tag all relevant items
5. Snowball: check references of key papers (backward) and citing papers (forward)
```

### Phase 3: Source Evaluation (20% of effort)

Rate each source on a standardized evidence hierarchy:

```
Level 1: Systematic reviews and meta-analyses
Level 2: Randomized controlled trials / controlled experiments
Level 3: Cohort studies / quasi-experimental designs
Level 4: Case-control studies / cross-sectional surveys
Level 5: Case reports / case series / expert opinion
Level 6: Anecdotal evidence / grey literature without methodology
```

**Credibility checklist per source**:

```markdown
□ Author credentials and affiliation
□ Publication venue (impact factor, peer-review process)
□ Methodology transparency (can you replicate it?)
□ Sample size and representativeness
□ Conflict of interest disclosure
□ Recency (is the data still relevant?)
□ Citation count and reception (supportive vs. critical citations)
□ Consistency with other sources (does it converge or contradict?)
```

### Phase 4: Evidence Synthesis (30% of effort)

Organize findings into structured artifacts:

#### Evidence Matrix

| Finding | Source(s) | Evidence Level | Strength | Notes |
|---------|-----------|---------------|----------|-------|
| LLMs improve code quality by 20-40% | [A], [B], [C] | Level 2-3 | Strong (convergent) | Effect varies by task complexity |
| Developers trust AI suggestions less for security-critical code | [D], [E] | Level 4 | Moderate | Small sample sizes |
| No significant effect on debugging time | [F] | Level 2 | Weak (single study) | Contradicts [A] — needs reconciliation |

#### Contradiction Log

When sources disagree, document systematically:

```markdown
## Contradiction: Effect of X on Y

**Position A**: X increases Y (Smith 2023, Jones 2024)
  - Evidence: RCT with n=500, effect size d=0.4
  - Context: University students, controlled setting

**Position B**: X has no effect on Y (Lee 2024)
  - Evidence: Field study with n=1200, p=0.34
  - Context: Industry practitioners, naturalistic setting

**Resolution hypothesis**: The effect is moderated by expertise level.
  Position A's sample (students) shows the effect;
  Position B's sample (practitioners) does not.
  → Need: Study that measures expertise as a moderator.
```

#### Knowledge Map

Visualize the landscape of your findings:

```
Central Question
├── Sub-Q1: [Strong evidence — 8 sources, convergent]
│   ├── Finding 1.1 (Level 2, 3 sources)
│   ├── Finding 1.2 (Level 3, 2 sources)
│   └── Finding 1.3 (Level 4, 3 sources)
├── Sub-Q2: [Mixed evidence — 5 sources, 1 contradiction]
│   ├── Finding 2.1 (Level 2, 2 sources)
│   └── Finding 2.2 ⚠️ CONTRADICTED by Finding 2.3
├── Sub-Q3: [Weak evidence — 2 sources, emerging area]
│   └── Finding 3.1 (Level 5, 2 sources)
└── Unexpected: [Theme that emerged during research]
    └── Finding 4.1 (Level 3, 1 source) → needs further investigation
```

### Phase 5: Deliverable Production (10% of effort)

Compile findings into the target deliverable format:

**For a Literature Review**:
1. Organize by themes (not chronologically)
2. Synthesize across sources (not paper-by-paper summaries)
3. Identify gaps explicitly ("No studies have examined...")
4. State implications for your research

**For a State-of-the-Art Report**:
1. Current landscape with taxonomy
2. Key advances and timelines
3. Open problems and active debates
4. Future directions with evidence basis

**For a Policy Brief**:
1. Executive summary (1 paragraph)
2. Evidence summary (1-2 pages)
3. Policy options with trade-offs
4. Recommended action with justification

## Iteration Protocol

Deep research is inherently iterative. After Phase 4, reassess:

```
After synthesis:
  □ Are all sub-questions adequately answered?
  □ Are there new sub-questions that emerged?
  □ Are there critical gaps requiring additional search?
  □ Are contradictions resolved or at least documented?

If gaps remain:
  → Return to Phase 2 with refined queries
  → Maximum 3 iteration cycles before declaring scope complete
  → Document what remains unknown (future work)
```

## Quality Indicators

A well-executed in-depth investigation should demonstrate:

- **Breadth**: Multiple source tiers consulted (not just Google Scholar)
- **Depth**: Key papers read in full, not just abstracts
- **Rigor**: Evidence levels assessed, contradictions documented
- **Transparency**: Search strategy reproducible, decisions justified
- **Currency**: Most recent relevant work included
- **Balance**: Competing viewpoints represented fairly

## References

- Petticrew, M., & Roberts, H. (2006). *Systematic Reviews in the Social Sciences*. Blackwell.
- Grant, M. J., & Booth, A. (2009). "A typology of reviews." *Health Information & Libraries Journal*, 26(2), 91-108.
- Snyder, H. (2019). "Literature review as a research methodology." *Journal of Business Research*, 104, 333-339.

版本历史

  1. v2026.07 GitHub ↗
    原版更新说明

    ## Highlights - **First tagged release.** Every gate is green: `make check` (validate + 183 unit tests + eval-harness + numeric benchmarks), validate-catalog, quality-evals, and OpenSSF Scorecard on `main`. - **Rigor double loop at 13 method families.** Each family closes taxonomy tag → eval scenario (does the agent avoid the method's known traps?) → deterministic numeric benchmark (does it recover the true effect?). New this cycle: heterogeneous treatment effects (CATE) and quantile treatment effects (QTE). - **Paper-WorkFlow flagship at 29/29 executable rigor gates**, now merged to its `main` with a drift-gated README badge — the credibility artifact is checked by CI, so it can never silently go stale ([RIGOR.md](https://github.com/brycewang-stanford/Paper-WorkFlow/blob/main/RIGOR.md)). - **1,150 vendored skills / 69 collections**, all license-audited, security-scanned, tracked in a machine-readable catalog with CI-enforced freshness, and importable as a single whole-repo router skill. - **Hardened weekly upstream syncs**: synced trees must pass the full gate before merge; this cycle both sync PRs were verified, patched (StatsPAI's SkillOpt execution-gate card restored), and merged. - **First-party plugin marketplace** (Python / R / Stata empirical-analysis plugins + guide) shipped in the previous cycle, now under a tagged version. ## Known Follow-Ups - Ecosystem benchmark comparison (AERS vs econometrics agents on the shared benchmark suite) — [PLAN-2026-07 week 3](https://github.com/brycewang-stanford/Auto-Empirical-Research-Skills/blob/main/docs/PLAN-2026-07.md). - Decomposition (Oaxaca–Blinder) and shift-share (Bartik) method families (weeks 3–4). - GitHub Pages catalog site, README demo screenshots, five-language README consistency script (week 4). ## Catalog - Top-level collections: **69** - `SKILL.md` files: **1150** - Cataloged agent tools: **335** ([`tools/CATALOG.md`](../tools/CATALOG.md)) ## Methodological rigor - Method families in the coverage map: **13** ([`RIGOR_COVERAGE.md`](RIGOR_COVERAGE.md)) - Eval scenarios: **26** (138 rubric items, 122 auto-checkable) ([`../eval-harness/`](../eval-harness/README.md)) - Benchmark tasks: **13** (70 gold checks, 58 required) ([`../benchmark/`](../benchmark/README.md)) ### Method families (skills tagged) - `iv`: 27 - `rdd`: 25 - `panel-fe`: 19 - `staggered-did`: 17 - `bayesian`: 13 - `did`: 10 - `event-study`: 10 - `matching`: 10 - `synthetic-control`: 10 - `cate`: 8 - `dml`: 8 - `survival`: 3 - `quantile`: 2 ### Eval scenarios by category - `causal-identification`: 14 - `citation-hygiene`: 1 - `reproducibility`: 2 - `research-integrity`: 1 - `runtime-safety`: 1 - `statistical-validity`: 2 - `writing-compliance`: 3 - `writing-style`: 2 ## License buckets - AGPL-3.0: 1 - Apache-2.0: 1 - CC-BY-NC-4.0 (non-commercial): 1 - CC-BY-SA-4.0: 1 - CC-BY-SA-4.0 (repository default): 4 - GPL-3.0: 1 - MIT: 35 - MIT Non-Commercial: 1 - Mixed (per-skill; see folder): 2 - UNKNOWN - check upstream: 22 ## Quality gates - `make check` — catalog/provenance/audit/eval freshness, repo hygiene, Paper-WorkFlow gates, stdlib unit tests, eval-harness lint + graded smoke, and the strict numeric benchmark (reference-gold drift fails the build). - External links and OpenSSF Scorecard run on their scheduled workflows.

当前收录版本

0.0.1

核对来源版本 ↗

质量评测

满分 100。前四项来自 DeepSkill 收录的质量报告。

场景价值92/100
评分依据

提供完整的5阶段深度调研框架,适用于学术研究、政策分析等高价值场景

触发精准95/100
评分依据

frontmatter 中 keywords 覆盖全面且精准

结构规范93/100
评分依据

5个Phase划分清晰,每个阶段有明确目标和产出

执行可靠94/100
评分依据

方法论基于权威参考文献,来源可靠

证据可验 (前四项均值)94/100
计算方式

取前四项原始分值的算术平均,四舍五入取整。可通过收录报告核对各项评分与说明,运行记录需单独查证。

权限与使用前检查

根据当前主文档自动整理,命中内容可能是示例或防护说明;不是实际权限声明,也未覆盖附件、外部依赖或运行行为。

  • 账号、密钥或授权

    使用专用、最小权限账号,不把密钥粘贴到公开内容或命令日志;核对接收数据的服务与租户。

    主文档提及 · 第 89 行

本站展示提示不会授予权限。涉及文件改动、密钥、外部发送或费用时,仍由你使用的工具按需确认。

安全评级与访问范围

risk · 静态证据核验,不代表运行安全保证。

使用前请由宿主核对并授权实际文件读写、外部服务和联网目的地;本站的核验不会授予任何运行权限。

来源与使用声明 SkillPi 的部分技能资料来自公开来源或创作者提交。第三方内容的权利归各自权利人所有。安装或运行前,请查看原站的许可、权限要求和使用风险。

来源与使用

请按详情页的作者、原始来源和许可证核对使用条件,运行前检查技能需要的权限。目录收录不等于已验证安装或安全性。