Agent skill
global-intelligence-search
企业级全局情报搜索网关。支持 PubMed 医学文献检索和通用网页抓取。这是 Investigator Agent 唯一合法的外部信息获取武器。
Install this agent skill to your Project
npx add-skill https://github.com/timoxue/LingNexus/tree/main/skills
SKILL.md
全局情报搜索网关 (Global Intelligence Search)
核心能力
这是一个三层解耦的企业级搜索架构：
- L0 网关层：智能路由 + 兜底容错
- L1 引擎层：医疗数据库引擎 + 浏览器引擎
- L2 解析层：HTML 清洗 + 文本提取
使用方法
命令格式
python skills/global_search_skill.py "<query>" "<domain>"
参数说明
query: 搜索关键词（PubMed）或目标 URL（通用网页）domain: 搜索域，必须为以下之一：pubmed- 检索 PubMed 医学文献数据库general_web- 抓取通用网页内容
使用场景
场景 1：检索医学文献
当你需要查阅严谨的医学文献、临床试验数据、药物研发信息时：
python skills/global_search_skill.py "PROTAC BRD4 degradation" "pubmed"
python skills/global_search_skill.py "molecular glue CDK inhibitor" "pubmed"
python skills/global_search_skill.py "靶向蛋白降解 LYTAC" "pubmed"
返回内容：
- PMID（PubMed ID）
- 文章标题
- 摘要（自动截断至 500 字符）
- 最多返回 10 条结果
场景 2：抓取开源网页
当你需要从公开网站获取数据（专利数据库、公司官网、新闻报道）时：
python skills/global_search_skill.py "https://clinicaltrials.gov/study/NCT12345678" "general_web"
python skills/global_search_skill.py "https://www.fda.gov/drugs/new-drugs" "general_web"
返回内容：
- 清洗后的纯文本（移除 script/style/nav 等非正文标签）
- 自动截断至 8000 字符
- 15 秒超时保护
容错保障
所有三层（L0/L1/L2）均采用极致容错设计：
- ✅ 网络超时自动捕获
- ✅ 403/反爬自动降级
- ✅ 解析失败返回错误字符串
- ✅ 绝不向上抛出异常导致容器崩溃
环境要求
Python 依赖
pip install beautifulsoup4 biopython
环境变量
export NCBI_EMAIL="your_email@example.com" # PubMed API 必需
示例输出
PubMed 检索示例
=== PubMed 检索结果 ===
关键词: PROTAC BRD4
[1] PMID: 12345678
标题: Targeted protein degradation by PROTACs
摘要: Proteolysis-targeting chimeras (PROTACs) are heterobifunctional molecules...
[2] PMID: 87654321
标题: BRD4 degradation in cancer therapy
摘要: BRD4 is a key epigenetic reader protein...
网页抓取示例
=== 网页抓取结果 ===
URL: https://example.com
Clinical Trial Information
Study Title: Phase I Study of XYZ-123
Status: Recruiting
Condition: Solid Tumors
Intervention: XYZ-123 (PROTAC)
...
严格使用规则
⚠️ 这是 Investigator Agent 唯一合法的外部信息获取工具
-
必须根据任务精准选择 domain：
- 医学文献 →
pubmed - 网页数据 →
general_web
- 医学文献 →
-
禁止行为：
- ❌ 不得使用其他未授权的网络工具
- ❌ 不得对返回结果进行总结或改写
- ❌ 不得过滤或删除原始信息
-
并发执行：
- 支持多任务并发调用
- 每个任务独立超时控制
- 互不阻塞
Recommended Agent Skills
Expand your agent's capabilities with these related and highly-rated skills.
mattpocock/skills
edit-article
Edit and improve articles by restructuring sections, improving clarity, and tightening prose. Use when user wants to edit, revise, or improve an article draft.
mattpocock/skills
migrate-to-shoehorn
Migrate test files from `as` type assertions to @total-typescript/shoehorn. Use when user mentions shoehorn, wants to replace `as` in tests, or needs partial test data.
mattpocock/skills
git-guardrails-claude-code
Set up Claude Code hooks to block dangerous git commands (push, reset --hard, clean, branch -D, etc.) before they execute. Use when user wants to prevent destructive git operations, add git safety hooks, or block git push/reset in Claude Code.
mattpocock/skills
obsidian-vault
Search, create, and manage notes in the Obsidian vault with wikilinks and index notes. Use when user wants to find, create, or organize notes in Obsidian.
mattpocock/skills
setup-pre-commit
Set up Husky pre-commit hooks with lint-staged (Prettier), type checking, and tests in the current repo. Use when user wants to add pre-commit hooks, set up Husky, configure lint-staged, or add commit-time formatting/typechecking/testing.
mattpocock/skills
scaffold-exercises
Create exercise directory structures with sections, problems, solutions, and explainers that pass linting. Use when user wants to scaffold exercises, create exercise stubs, or set up a new course section.
Didn't find tool you were looking for?