AgentSkillsCN

perplexity-deep-search

借助 Perplexity AI 进行深度研究,生成涵盖引文与分析的全面报告。

SKILL.md
--- frontmatter
name: perplexity-deep-search
version: 1.0.0
description: Deep research using Perplexity AI. Comprehensive reports with citations and analysis.
metadata:
  openclaw:
    emoji: 🔍
    category: research

Perplexity Deep Search

Deep research using Perplexity AI. Comprehensive reports with citations and analysis.

When to Use

Use CaseExample
Market research"Competitive analysis of AI coding assistants"
Technology evaluation"Compare vector databases for production use"
Trend analysis"State of MCP ecosystem 2025"
Due diligence"Company X funding, team, product overview"
Learning"Explain quantum computing to a programmer"

vs Other Research Tools

ToolBest ForDepth
PerplexityQuick, cited answersMedium
Pokee Deep ResearchComprehensive reportsDeep (25+ min)
Web SearchSpecific factsShallow
arXiv WatcherAcademic papersDeep

Workflow

1. Targeted Query

Be specific:

code
"Compare Supabase vs Firebase for real-time apps in 2025. 
Focus on: scalability, pricing, self-hosting options."

2. Follow-Up Questions

Drill deeper:

code
"What are the specific limitations of Firebase's real-time 
sync at 10k+ concurrent users?"

3. Synthesis

Combine multiple sources:

code
"Based on the above, which is better for a startup with 
< 1000 users planning to scale to 100k?"

Output Format

Request structured output:

markdown
## Summary
2-3 sentence overview

## Key Findings
- Point 1 (with citation)
- Point 2 (with citation)

## Comparison Table
| Criteria | Option A | Option B |
|----------|----------|----------|

## Recommendations
1. Best for X
2. Best for Y

## Sources
- [Title](URL)

Automation

json
{
  "name": "weekly-tech-digest",
  "schedule": {"kind": "cron", "expr": "0 9 * * 1"},
  "payload": {
    "kind": "agentTurn",
    "message": "Use perplexity to research: 'AI agent frameworks 2025'. Focus on new releases, emerging standards, key players."
  },
  "sessionTarget": "isolated",
  "notify": true
}

Best Practices

  1. Be specific — Narrow queries get better results
  2. Request citations — Verify claims
  3. Follow up — One query rarely tells the full story
  4. Save findings — Add to MEMORY.md for future reference
  5. Cross-check — Confirm critical facts with primary sources