The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill's primary function is to aggregate and summarize news from external RSS feeds and websites. This creates a surface for indirect prompt injection attacks, where a malicious or compromised news source could include hidden instructions designed to influence the agent's behavior during the briefing generation phase.\n
Ingestion points: scripts/fetch_news.py and scripts/rss_parser.py ingest data from multiple external domains including 36kr.com, api-one.wallstcn.com, i.news.qq.com, and several RSS feeds.\n
Boundary markers: There are no explicit boundary markers or instructions to the agent to ignore embedded commands within the fetched data.\n
Capability inventory: The skill possesses the capability to execute its own Python scripts which perform network GET requests and write output files to the local file system.\n
Sanitization: The scripts use BeautifulSoup to extract plain text from HTML content, which removes markup but does not filter for instructional language that might manipulate the agent.\n- [COMMAND_EXECUTION]: The skill relies on the agent executing local Python scripts (scripts/fetch_news.py, scripts/daily_briefing.py) to perform its core operations. These scripts have the permission to make network requests and modify the local file system by saving reports.\n- [EXTERNAL_DOWNLOADS]: The skill fetches data from several external news platforms and RSS feeds, such as 36kr, Wall Street CN, and Tencent News. While these are well-known and reputable Chinese news services, they are external dependencies for the content.\n- [COMMAND_EXECUTION]: In scripts/rss_parser.py, the fetch_rss_feed function uses verify=False in the requests.get call. This explicitly disables SSL certificate verification, making the connection vulnerable to man-in-the-middle (MITM) attacks.

skill-news-intelligence