Rss Aggregator

自动读取配置的 RSS 链接，抓取并合并多源报道，去重历史推送，生成无 Emoji、高信息密度的 Markdown 简报。

Security Scan

Skill mostly matches an RSS-aggregation description but has inconsistencies (undeclared local files, explicit instruction to hide provenance) that warrant caution before install.

Purpose & Capability

The skill's stated purpose (fetch RSS, dedupe, produce Markdown briefs) aligns with the instructions to read rss_sources.txt and fetch articles. However the registry metadata declared no required config paths or files, while SKILL.md expects a local rss_sources.txt and a pushed_history.log for deduplication—this mismatch between declared requirements and actual runtime file access is an inconsistency.

Instruction Scope

Runtime instructions explicitly tell the agent to read rss_sources.txt, fetch and parse full web articles, and read/append pushed_history.log. They also mandate 'hide data source' (never display RSS feeds or original outlet names). Reading/writing a local history file and suppressing provenance are legitimate for brief generation but are sensitive: the skill can erase or omit source attribution and persist a local log without declaring those files. The instructions give the agent broad discretion to judge 'is_abuse=true' and to merge cross-source content—this is content-sensitive and may cause omitted provenance or untrusted merging decisions.

Install Mechanism

No install spec or code files are present; it's instruction-only so nothing is written to disk by an installer. This is lower risk from an installation standpoint.

Credentials

The skill requests no environment variables or external credentials, which is proportionate. However, it accesses local files (rss_sources.txt is present; pushed_history.log is referenced but not present in manifest) despite metadata claiming no required config paths—this undeclared file access should be made explicit.

Persistence & Privilege

always is false and the skill does not request elevated platform privileges. It does instruct writing to its own pushed_history.log (persistence limited to its own file), which is normal for incremental push functionality. There is no indication it modifies other skills or global agent configuration.

What to consider before installing

What to consider before installing: - The skill will read rss_sources.txt and expects a pushed_history.log (for dedupe) even though the registry metadata didn't declare config files—confirm where these files will live and that you trust that location. - The skill requires network access to fetch article pages and will parse full article content; run it in a sandbox or with limited network rights if you need containment. - It explicitly instructs the agent to hide source attribution (never show RSS feeds or outlet names). If provenance is important for your use case (audit, credibility), this behavior is undesirable. - The instructions give the agent discretion to judge and drop 'abusive' content and to merge multi-source reports—review sample outputs to ensure merging/omission logic meets your standards. - Because this is instruction-only with no code, review the rss_sources.txt contents and consider creating/inspecting the pushed_history.log beforehand; add monitoring of that file so you know what has been recorded. - If you want stronger assurance, request the skill author to: (1) declare config paths in metadata, (2) document exact file locations and formats for pushed_history.log, and (3) remove or explain the requirement to hide original sources.

智能 RSS 聚合与增量推送 (Smart RSS Aggregator)

技能描述

自动读取 RSS 链接列表，抓取并阅读正文。通过跨源事件合并与严格的历史日志核对，实现每日多次运行时的“纯增量”推送，最终生成高信息密度的纯文本简报。

执行指令 (Workflow)

当触发此技能时，请严格按以下顺序执行：

1. 资源读取与抓取 读取配置文件中的有效 RSS 链接，获取最新文章并通读网页核心内容。
2. 内容过滤 (质量审查与跨源合并)
   • 内容审查：对抓取的内容进行初步评估，若检测到垃圾广告、无意义水文或滥规言论（即内部判定 is_abuse=true），请直接丢弃。
   • 跨源合并：交叉分析通过审查的文章。若多篇报道指向同一事件，将其合并为一条综合资讯，提取各方补充细节，避免同一事件重复出现。
3. 严格核对记忆 (核心增量逻辑) 读取本地的 pushed_history.log 文件。对比当前准备推送的事件，彻底剔除所有在历史记录中已经存在的内容。 （注：确保每天 18 点运行时，不会推送当天 8 点已经推送过的文章。一旦某条全新资讯被选中推送，必须立即将其标题和链接追加记录到 pushed_history.log 中。）
4. 撰写与排版 将所有筛选出的全新资讯，严格按照下方的【输出规范】和【输出模板范例】进行排版。

输出规范 (Strict Formatting Rules)

必须严格遵守以下约束，任何情况下均不可违反：

• 头部日期：推送的最顶部必须有且仅有一行总标题：今日AI资讯 (YYYY-MM-DD)，按当天实际日期动态填充。
• 区块结构：每条资讯严格按照 标题： -> 摘要： -> 链接： 的顺序排列，且每个字段之间必须保留一个空行。
• 深度摘要：摘要部分需要详尽且专业（类似专业科技媒体的深度快讯）。要求包含具体数据（如时间、算力、参数量等）、核心动作和深层影响，直接陈述事实，不加任何主观废话。
• 隐藏数据源：绝对不要在任何地方展示 RSS 订阅源或原始媒体的名称。
• 纯净排版：严禁使用任何 Emoji 表情符号，仅使用纯文本及基础的 Markdown 换行。

输出模板范例

请严格套用以下模板生成最终推送（除了填充内容外，不要改变任何换行和标点符号）：

今日AI资讯 (YYYY-MM-DD)

标题： [此处填写合并后的事件标题]

摘要： [此处撰写详尽的核心摘要。多源合并时需综合各方要点，提取最核心的数据、过程和结论。]

链接： [原文链接 1] [原文链接 2 (若有多个链接直接换行展示)]

标题： [此处填写下一条独立事件标题]

摘要： [下一条事件的详细摘要...]

链接： [原文链接]