The Agent Skills Directory

[DATA_EXFILTRATION]: The skill fetches data from multiple external sources including RSS feeds, Hacker News API (Algolia), and HuggingFace API. While these are reputable services, it involves outbound network requests to aggregate content.
[EXTERNAL_DOWNLOADS]: The script scripts/fetch_ai_news.py dynamically fetches README files from raw.githubusercontent.com based on repositories found on the GitHub Trending page. This content is used to supplement the newsletter summaries.
[PROMPT_INJECTION]: The skill is susceptible to indirect prompt injection due to its core functionality of summarizing external web content.
Ingestion points: The skill ingests text from 20+ external RSS feeds, Hacker News stories, and GitHub README files (processed in fetch_ai_news.py).
Boundary markers: There are no explicit delimiters or instructions provided in the workflow to the subagents to ignore potential instructions embedded within the fetched news data.
Capability inventory: The skill has the capability to write files to the local disk (reports/ directory) and execute local scripts for rendering HTML and taking screenshots.
Sanitization: Basic sanitization is performed by stripping HTML tags using BeautifulSoup, but the remaining text content is not analyzed for malicious instructions before being passed to subagents for summarization.

ai-daily-newsletter