| name | in-depth-research |
| version | 1.0.0 |
| description | 深度研究技能的红龙定制包装入口。检测到 HONGLONG-OVERRIDE.md 时优先使用定制内容。 |
| triggers | ["深度研究","深度调研","in-depth research"] |
技能路由入口 — in-depth-research
本文件为 in-depth-research 技能的路由入口。
不随社区技能更新覆盖,保持红龙定制逻辑。
路由逻辑
AI Agent 执行以下检测:
本文件所在目录是否存在 HONGLONG-OVERRIDE.md?
│
├─ 是 → 读取 HONGLONG-OVERRIDE.md 内容作为完整技能指令
│ (跳过以下原始社区技能内容)
│
└─ 否 → 读取下方原始社区技能内容执行
为什么这样设计:HONGLONG-OVERRIDE.md 是独立文件,社区技能更新只覆盖 SKILL.md,
不覆盖 HONGLONG-OVERRIDE.md,红龙定制内容得以保留。
原始社区技能内容
以下为原始 clawdhub 技能内容(仅当 HONGLONG-OVERRIDE.md 不存在时使用)
In-Depth Research
Multi-source research skill that validates findings across sources and delivers decision-ready intelligence.
Research Methodology
Phase 1: Source Identification
- Define research scope and key questions
- Identify authoritative primary sources (industry reports, government data, academic papers)
- Map secondary sources (news, blogs, company filings)
Phase 2: Parallel Collection
- Execute searches across all identified sources simultaneously
- Document source quality and recency
- Flag contradictions immediately
Phase 3: Triangulation
- Cross-validate key facts across 3+ independent sources
- Weight findings by source reliability
- Calculate confidence scores for each finding
Phase 4: Synthesis
- Structure output by research question
- Distinguish facts from interpretations
- Highlight areas of uncertainty
Output Standards
| Standard | Requirement |
|---|
| Source count | Minimum 5 independent sources |
| Recency | No data older than 24 months without caveat |
| Attribution | Every claim traced to source |
| Confidence | High/Medium/Low/Unknown for each finding |