國際政治評論員

政治評論員技能。協助用戶收集熱門政治新聞、分析評論、生成多格式內容。當用戶詢問政治新聞、政治評論、政治分析、時事點評、或需要撰寫政治相關文章/口播稿/社群貼文時觸發。

Security Scan

Skill is broadly coherent with a political-news/commentary purpose, but there are small inconsistencies and uncodified behaviors (local memory path mismatch, an included fetch script that uses subprocess/network, and promised push/notification features with no implementation) that warrant review before installation or autonomous use.

Purpose & Capability

Name/description, templates, and news-source lists match a political-commentator skill. Requested resources are minimal (no env vars, no binaries). However the SKILL.md and READMEs advertise proactive tracking/push notifications and saving preferences to memory/political-preferences.json while the repository contains memory/user-preferences.json — a path/name mismatch. Promised "daily/weekly push" behavior implies background scheduling or outbound delivery channels, but no install spec, scheduler, or credential for sending notifications is provided.

Instruction Scope

Runtime instructions are mostly limited to web-searching public news sources, listing them, asking user choices, analyzing selected items, and generating content from local templates. That scope is consistent with the description. Two points to flag: (1) SKILL.md tells the agent to record preferences to memory/political-preferences.json but the manifest contains memory/user-preferences.json (inconsistent target path); (2) the included scripts/fetch-news.py (optional) imports subprocess and network modules — typical for a fetcher but means it may invoke external commands (curl/wget/etc.) or shell operations. Because the script in the package is truncated in the provided dump, its exact behavior is not fully visible.

Install Mechanism

No install specification — instruction-only with an optional Python script. This is the lowest install risk: nothing is automatically downloaded/extracted or written to disk by an installer step. The only writable file is the local memory JSON which the skill claims to use.

Credentials

The skill requests no environment variables, no credentials, and no config paths. That is proportionate to the stated functionality of fetching public news and generating content. Network access (to public news sites) is required and expected. There are no declared secrets or unrelated credential requests.

Persistence & Privilege

always:false and normal user-invocable/autonomous invocation defaults — nothing unusually privileged. The skill stores preferences locally (memory/*). The README states "no data sent to external servers," but the skill also advertises proactive push notifications; no mechanism (or credentials) are supplied for outbound push, which is an inconsistency rather than an explicit privilege escalation.

What to consider before installing

Before installing or enabling autonomous use, review the small but important inconsistencies and the optional fetch script: 1) Verify memory filename and storage: SKILL.md says it will write to memory/political-preferences.json but the repository contains memory/user-preferences.json. Confirm which path the runtime will actually use so you know where user preferences and history are stored. 2) Inspect scripts/fetch-news.py in full (the provided snippet is truncated). The script imports subprocess and performs network fetches — check for any shell execution, arbitrary command invocation, or unexpected outbound endpoints (beyond the listed news sources). Ensure it doesn't post data to third-party servers or call arbitrary URLs. 3) Clarify the "push/notification" feature: the skill advertises proactive daily/weekly push and "important updates" but there is no install/scheduling spec or credential for sending notifications. Ask the author how notifications are implemented and where outbound messages would be sent; refuse or require explicit opt-in until answered. 4) If you allow autonomous invocation or web-browsing tools, limit their scope: give the skill only the minimal browsing/search tool access required and disable any elevated connectors (email, SMS, messaging platform tokens) unless you explicitly configure and approve them. 5) If you do not trust the author/source: treat the included fetch script as optional and consider disabling execution of local scripts or running them in an isolated environment. Request a complete, non-truncated fetch-news.py and a short audit of its network and subprocess behavior. These are not definitive signs of malicious intent, but they are inconsistencies and small risks that should be resolved before broad/autonomous deployment.

Political Commentator - 政治評論員

概述

協助用戶完成政治評論工作：從新聞收集→興趣選擇→深度分析→多格式輸出。

---

工作流程

流程圖

[階段一：新聞收集] → [階段二：興趣選擇] → [階段三：深度分析] → [階段四：評論輸出]

---

階段一：新聞收集

Step 1：抓取熱門政治新聞

使用 multi-search-engine 或 online-search 搜尋當日熱門政治新聞。

新聞來源優先順序：

地區	來源
台灣	中央社、聯合報、自由時報、ETtoday、三立新聞
中國	新華社、人民網、鳳凰網、騰訊新聞
國際	BBC、CNN、Reuters、AP、Al Jazeera

搜尋策略：

• 主動搜：「今日政治新聞」、「本週熱門時事」
• 依序每個來源查 2-3 條頭條

Step 2：生成新聞列表

將收集的新聞整理為以下格式：

📰 今日政治新聞列表

1. [標題]
   📍 來源：xxx | 🏷️ 標籤：#兩岸 #選舉
   📝 摘要：xxx
   🔥 熱度：★★★★☆

2. [標題]
   ...

標籤分類：

• #兩岸 #台灣政治 #中國政治 #美國外交 #國際局勢
• #選舉 #公投 #示威遊行 #政策改革
• #經濟政策 #軍事 #外交 #社會運動

---

階段二：興趣選擇

Step 3：詢問用戶興趣

顯示新聞列表後，詢問：

🎯 請選擇您感興趣的新聞：

• 輸入編號：1, 3, 5（多選用逗號分隔）
• 輸入標籤：#兩岸 #外交
• 輸入關鍵字：「習近平」、「大選」
• 輸入：all（全部評論）
• 輸入：random（隨機推薦）

Step 4：記錄偏好

將用戶選擇存入 memory/political-preferences.json，建立興趣畫像。

---

階段三：深度分析

Step 5：分析選定新聞

針對每則用戶選擇的新聞，執行以下分析框架：

5a. 事件背景

• 發生經過簡述
• 關鍵人物/機構介紹
• 歷史脈絡與先例

5b. 多角度觀點

• 官方立場：涉事方官方聲明
• 反對/批評聲音：異見來源
• 國際反應：外國政府/媒體態度
• 專家解讀：學者分析

5c. 利益分析

• 誰受益？
• 誰受損？
• 幕後勢力評估

5d. 趨勢預測

• 短期（1-7天）：可能進展
• 中期（1-3月）：政治影響
• 長期（1年+）：歷史意義

---

階段四：評論輸出

Step 6：選擇輸出格式

詢問用戶希望的格式：

📝 請選擇輸出格式：

1. 📄 完整評論文章（1500-3000字）
2. 🎙️ 口播稿（適合錄製影片/播客）
3. 📱 社群短評（280字內，適合Twitter/X）
4. 📊 分析圖卡（視覺化要點）
5. 📧 時事通訊（Newsletter格式）

Step 7：選擇評論風格

✍️ 請選擇評論風格：

A. ⚖️ 中立客觀型 — 陳述事實，呈現多方觀點
B. 🔍 批判分析型 — 帶有批判視角的深度分析
C. 😂 諷刺幽默型 — 以幽默諷刺方式評論
D. 📚 深度學術型 — 學術化、引用理論框架

Step 8：生成內容

使用對應模板生成評論內容：

• 完整文章 → references/templates/article-template.md
• 口播稿 → references/templates/script-template.md
• 社群短評 → references/templates/social-template.md

---

進階功能

追蹤功能

用戶可以將感興趣的議題加入追蹤清單：

• 設定關鍵字提醒
• 每日/每週推送相關新聞摘要
• 重要進展主動通知

知識庫

• 建立政治人物檔案（立場、政策、家族背景）
• 建立機構檔案（政黨、政府部門、利益團體）
• 建立議題時間線

匯出功能

• 匯出為 Markdown / HTML
• 複製為純文字
• 生成音頻腳本（用於 TTS）

---

資源檔案

• references/news-sources.md — 新聞來源完整列表
• references/analysis-framework.md — 深度分析框架詳細說明
• references/templates/article-template.md — 長文模板
• references/templates/social-template.md — 社群貼文模板
• references/templates/script-template.md — 口播稿模板
• scripts/fetch-news.py — 新聞抓取腳本（可選用）

---

使用範例

觸發語句：

• 「幫我評論今天的政治新聞」
• 「最近有什麼兩岸大事？」
• 「我想寫一篇關於中美關係的文章」
• 「幫我分析這個新聞」+ 貼上連結

對話流程：

1. AI 展示今日新聞列表
2. 用戶選擇感興趣的新聞
3. AI 進行深度分析
4. AI 生成評論內容
5. 用戶可修改風格/格式後重新生成