AgentSkillsCN

Research

通过统筹exa-search(网络搜索)、github-quality-search(代码搜索)与hn-research(趋势/情绪分析),开展多源深度研究。适用于需要综合网络、开源与社区资源,以获得全面解答的场景。

SKILL.md
--- frontmatter
description: Deep multi-source research by orchestrating exa-search (Web), github-quality-search (Code), and hn-research (Trends/Sentiment). Use for comprehensive answers requiring synthesis from web, open source, and community.

Research Specialist

A meta-skill for conducting deep, multi-source research by orchestrating exa-search (Web), github-quality-search (Code/Libraries), and hn-research (Trends/Sentiment). Use this when the user needs a comprehensive answer that requires synthesizing information from the web, open source ecosystem, and community discussions.

Tools & Capabilities

1. Web Research (Exa)

  • Source: exa-search
  • Use for: Technical documentation, blogs, official sites, comparative analysis, and solving general queries.
  • Action: Use exa-search to gather factual ground truth.

2. Ecosystem Research (GitHub)

  • Source: github-quality-search
  • Use for: Finding libraries, tools, frameworks, and checking the health/activity of solutions.
  • Action: Use github-quality-search to find the best tools for the job.

3. Community Research (Hacker News)

  • Source: hn-research
  • Use for: Gauging sentiment, finding hidden gems, understanding historical context, and "real world" usage reports.
  • Action: Use hn-research to see what real engineers think.

Workflow

  1. Analyze Request: Determine which vectors (Web, Code, Community) are relevant.
  2. Parallel Execution:
    • If looking for a solution: Check GitHub for tools + Exa for reviews.
    • If exploring a topic: Check Exa for facts + HN for opinions.
    • If debugging: Check Exa for fixes + GitHub issues.
  3. Synthesis: Combine findings into a structured report.
    • Summary: High-level answer.
    • Deep Dive: Evidence from Exa.
    • Tools: Recommendations from GitHub.
    • Community Consensus: Vibe check from HN.

Guidelines

  • Cross-Reference: Don't rely on one source. If a library looks good on GitHub, check HN to see if people hate it in production.
  • Citations: Always attribute where the info came from (e.g., "According to the docs...", "HN users discuss...", "The repo has 5k stars...").
  • Synthesize, Don't Dump: Do not just list search results. Read them and write a coherent answer.