360 Secure SkillHub
360 安全技能中心

portal-news-triage

该 Skill 使用声明中的 portal MCP 服务检索、筛选并比较新闻记录;优先查看元数据与摘要,仅在必要时获取记录,并依据每条记录的许可说明和省略信息呈现结果,保留原始来源引用。适合主题调研与带出处的新闻摘要。

chinng-inta
chinng-inta
热度 7.65
最近更新 2026-09-06 05:39:20
分类:搜索依赖:MCP 服务应用场景:研究 / 摘要总结 / 报告撰写

安装技能

复制下面的话术发给你的 Agent,它会通过 secure-skills 检索并安装该技能。

执行命令安装此 skill: npx secure-skills-cli add news-triage
Metadata
name
news-triage
version
0.2.1
license
MIT-0
description
Research and compare news from the Chinng AI-Agent Portal. Use when finding relevant articles, investigating a topic, or preparing a sourced news summary.

News triage

Use the portal MCP server declared by this plugin.

Standalone install (no plugin): register the read-only MCP endpoint first — openclaw mcp add portal --transport http https://portal.chinng-lab-srv.dev/mcp. Other MCP clients: https://portal.chinng-lab-srv.dev/mcp.

  1. Discover candidates with the portal's listing or search tools. Inspect metadata and summaries before requesting bodies.
  2. Select only the records needed for the task, then retrieve those records individually.
  3. Respect each record's license_note and omission metadata. Never reconstruct content withheld by the portal.
  4. Cite the portal record and retain its original source link when presenting findings.

What a news record actually contains

Most external news is link-only. Those records carry no summary at all — only the title, the metadata fields, and the links — and their summary_omitted_reason says why. Triage them on title, tags, keywords_en, entities_en, and numeric_facts, which remain available, and present them as dated pointers.

A full news record's body is a wrapper: the title, the summary repeated as TL;DR, the tags as Key Points, then an empty Details section. There is no article text behind it. Retrieving the body of a news record therefore rarely adds anything the listing did not already return, so retrieve one only when the summary is genuinely needed and was not present in the listing.

Treat entities as a triage hint whose precision varies by source, and ignore related_auto entirely.

Keep discovery token-light: list or search first, retrieve bodies only when they materially improve the answer.