AgentSkillsCN

research

围绕某一主题展开深度研究，打造可供日后查阅的持久性文档。适用于技术决策、竞品分析，或面对复杂议题时使用。

SKILL.md
--- frontmatter
name: research
description: "Deep research on a topic, creating persistent documentation for future reference. Use for technology decisions, competitive analysis, or complex topics."
model: claude-opus-4-5-20251101
allowed-tools: Read, Write, Edit, Glob, Grep, WebSearch, WebFetch

/research

Deep research creating persistent documentation for future reference. Uses Ultrathink for comprehensive analysis.

Usage

bash
/research "authentication patterns for multi-tenant SaaS"
/research "real-time collaboration architectures"
/research "Jira vs Linear competitive analysis"
/research "WebSocket vs SSE trade-offs"

Output Locations

code
ideas/
├── [project]/notes/research/     # Project-specific
├── shared/docs/research/         # Cross-project
└── resources/research/           # General reference

Location rules:

  • •Project-specific: ideas/{project}/notes/research/
  • •Cross-project: shared/docs/research/
  • •General reference: resources/research/

Process

1. Check Existing Research

bash
grep -r "authentication" ideas/*/notes/research/
grep -r "authentication" shared/docs/research/

If found, ask: "Research exists. Update, view, or create new?"

2. Deep Research Phase

Invoke research-specialist agent:

Research scope:

  • •Official documentation (via Context7)
  • •Technical blogs and tutorials
  • •Stack Overflow (high-vote answers)
  • •GitHub discussions and issues
  • •Community forums
  • •Existing solutions/products

Volume: Read 20-30+ sources, distill to 3-5 pages

3. Create Research Document

Naming: lowercase kebab-case (e.g., auth-patterns-saas.md)

Structure:

markdown
---
created: YYYY-MM-DD
updated: YYYY-MM-DD
sources: NUMBER
related:
  - path/to/related/file.md
tags: ["primary-tag", "secondary-tag"]
---

# Research: [Topic]

## Executive Summary
[2-3 paragraphs of key findings]

## Detailed Findings

### [Subtopic 1]
[Analysis]

### [Subtopic 2]
[Analysis]

## Recommendations
[What to do based on research]

## Must-Read Resources
1. **[Title]** - [url]
   - Why: [Reason]
   - Key point: [Takeaway]

## Additional Resources
- [Title] - [url]

## Gotchas & Warnings
- [Warning 1]

## Decision Matrix (if comparing)

| Criteria | Option A | Option B |
|----------|----------|----------|
| Ease of use | ⭐⭐⭐ | ⭐⭐ |
| Scalability | ⭐⭐⭐⭐ | ⭐⭐⭐ |

4. Link to Related Work

If research relates to a spec or issue:

  • •Add reference in research doc
  • •Update spec/issue to reference research

When to Use

Good candidates:

  • •Technology decisions affecting architecture
  • •Competitive analysis for product positioning
  • •Complex topics with many options
  • •Best practices you'll reference multiple times
  • •Market research for product ideas

Not needed for:

  • •Quick one-off questions (just ask)
  • •Simple API lookups (use Context7)
  • •Project-specific debugging

Updating Research

If research exists but is outdated:

code
Options:
1. Update existing (add new findings)
2. Replace entirely (fresh research)
3. View existing
4. Cancel

Integration

Research documents are discovered by:

  • •/plan - Checks for relevant research before planning
  • •/spec - References research in technical notes
  • •Future sessions - Research persists for reference
code
/research "topic" → ideas/{project}/notes/research/topic.md
                           ↓
              Future work automatically references this