Agent Skills
› nicknisi/claude-plugins
› research
research
GitHub并行启动多个AI代理以进行全方位研究。支持quick、standard和extensive三种强度，涵盖Perplexity、Claude等五种代理类型。通过单消息批量调用任务工具执行，并在完成后综合结果、识别共识与矛盾点。
Trigger Scenarios
需要全面或深度调研复杂主题
用户指令中包含research关键词及强度参数
Install
npx skills add nicknisi/claude-plugins --skill research -g -y
SKILL.md
Frontmatter
{
"name": "research",
"description": "Launch multiple AI agents in parallel for comprehensive research coverage",
"disable-model-invocation": true
}
Research Mode
Launch multiple agents simultaneously for comprehensive research coverage.
Usage
/consultant:research quick What are the best testing frameworks for React?
/consultant:research standard How do modern authentication systems work?
/consultant:research extensive Deep dive into microservices architecture patterns
/consultant:research What is the current state of WebAssembly? (defaults to standard)
Research Intensities
| Intensity | Agents | Best For |
|---|---|---|
| quick | 5 (1 of each) | Simple queries, fast answers |
| standard | 15 (3 of each) | Comprehensive coverage |
| extensive | 40 (8 of each) | Deep-dive, exhaustive reports |
Agent Types
perplexity-researcher: Citations, current dataclaude-researcher: Built-in WebSearch, detailed analysiscodex-researcher: GPT-5 with high reasoninggemini-researcher: Multi-perspective synthesisgrok-researcher: Alternative LLM perspective
Execution
Parse $ARGUMENTS to extract:
- Intensity: First word if
quick,standard, orextensive(default:standard) - Topic: The research topic
Launch agents using multiple Task tool calls in ONE message.
IMPORTANT: All agents MUST be launched in a single message with multiple parallel Task tool calls.
After agents complete, synthesize results:
- Identify consensus points (multiple agents agree)
- Highlight unique insights from specific agents
- Note contradictions or disagreements
- Rate confidence based on corroboration
Version History
- c3164c9 Current 2026-07-24 21:10
