Dingtalk Ai Web Search

网页搜索与实时信息检索。需要查找最新资讯、实时数据、技术文档、热点事件，或用户说"搜一下"、"帮我查"、"查资料"、"web search"等时使用。基于小宿AI智能搜索，支持关键词搜索、时间范围过滤（一天/一周/一月/一年）、自定义结果数量、JSON 输出。

Security Scan

The skill's code and runtime instructions match its stated purpose (a web search front-end to a DingTalk MCP search gateway); it requires no unrelated credentials or installs, but it does persist a user-supplied MCP URL to disk so users should treat that URL as sensitive before saving.

Purpose & Capability

Name/description (web search via 小宿智能搜索 / DingTalk MCP) align with the included script and instructions. The script talks to an MCP streamable-http URL and performs tool listing and search calls — expected for this purpose.

Instruction Scope

SKILL.md instructs the agent to ask the user for an MCP configuration JSON from a DingTalk admin page and to run connectivity checks. This is within scope for a gateway-backed search tool, but it explicitly requests the user-provided MCP JSON/URL and instructs saving it; that JSON/URL may be sensitive (could embed private endpoints or tokens) so the behaviour should be treated as storing credentials-like data.

Install Mechanism

Instruction-only skill with a bash script; no install spec, no network downloads or package installs. Uses standard system tools (curl, jq, python) — proportional and expected.

Credentials

The skill declares no required environment variables or credentials. The only sensitive input is the MCP JSON supplied by the user at runtime; that is logically required to contact the configured MCP gateway and is consistent with the stated purpose.

Persistence & Privilege

Script persistently writes the MCP URL string to a local file (.mcp_url) next to the script. always:false and no modifications to other skills or system configs. Persisting the user-supplied gateway URL is normal for convenience but has privacy implications — the file may contain sensitive endpoint information and should be protected.

Assessment

This skill is coherent for its stated purpose, but before installing or using it: 1) Treat any MCP JSON/URL you paste in as sensitive — it can reveal private endpoints or tokens. Only provide a URL from a trusted DingTalk/organization source. 2) The script saves the URL to .mcp_url in the skill directory — check and secure that file (permissions, location) or avoid using --save if you prefer not to persist it. 3) Expect the skill to make network calls to the supplied URL on session start (--ping) and during searches. 4) If you are unsure about the MCP config contents, inspect it before passing it to the skill, or run the script in a sandboxed environment. 5) If you need stronger guarantees, request a version that uses a platform-managed secret store rather than writing plaintext to disk.

会话开始：连通性检查

每次新会话开始，先运行 --ping 检查，再执行实际搜索：

bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh --ping

根据结果处理：

结果	含义	处理方式
✓ 连通成功，可用工具: web_search	正常	直接开始搜索
错误：未配置 MCP URL	从未配置或配置丢失	向用户索取 MCP 配置（见下方）
连接失败: ...	URL 失效或网络问题	让用户重新获取配置并执行 --save

---

配置缺失时：向用户索取 MCP 配置

让用户打开以下页面，复制 MCP 配置 JSON 后提供给你：

https://mcp.dingtalk.com/#/detail?detailType=instanceMcpDetail&instanceId=78440

配置格式示例：

{
  "mcpServers": {
    "小宿智能搜索": {
      "type": "streamable-http",
      "url": "https://mcp-gw.dingtalk.com/server/..."
    }
  }
}

收到配置后，运行连通性测试并永久保存（将 JSON 作为字符串传入 -c）：

bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh --save -c '{"mcpServers":{"小宿智能搜索":{"type":"streamable-http","url":"<用户提供的URL>"}}}'

• 成功：输出 ✓ 连通成功，可用工具: ... 并提示已保存，后续无需再传 -c
• 失败：根据报错提示处理（配置格式错误 / 网络不通 / URL 无效）

---

快速参考

场景	命令
会话开始连通检查	bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh --ping
保存配置（需 -c 传入 JSON）	bash ... --save -c '<MCP JSON配置>'
普通搜索	bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "关键词"
限定时间范围	bash ... -q "关键词" -f oneWeek
返回更多结果	bash ... -q "关键词" -n 10
JSON 输出	bash ... -q "关键词" --json

---

基础用法

bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "Python asyncio 最佳实践"

---

时间筛选

# 一周内的最新文章
bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "LLM 评测基准 2025" -f oneWeek

# 一个月内
bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "AI 编程工具对比" -f oneMonth

可选值: noLimit（默认）| oneDay | oneWeek | oneMonth | oneYear

---

JSON 输出（便于程序处理）

bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "FastAPI 性能优化" -n 5 --json

输出格式：

[
  {
    "title": "页面标题",
    "url": "https://...",
    "snippet": "内容摘要...",
    "site": "网站名称",
    "published": "2025-01-01",
    "source": "小宿智能搜索"
  }
]

---

典型工作流

调研某个技术主题

bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "LangGraph checkpoint 持久化方案" -n 5

获取实时信息

# 近一周新闻
bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "OpenAI o3 发布" -f oneWeek -n 8

多轮深入调研

# 第一轮：宽泛了解
bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "Rust 异步运行时对比" -n 5
# 第二轮：聚焦具体问题
bash <CURRENT_SKILL_MD_PATH_DIR>/scripts/search.sh -q "tokio vs async-std 性能基准测试 2024" -f oneYear -n 5