The Agent Skills Directory

[EXTERNAL_DOWNLOADS] (LOW): The skill relies on external Python packages (feedparser, requests) and fetches data from a remote RSS feed at https://news.smol.ai/rss.xml.- [DATA_EXFILTRATION] (LOW): The script performing network operations targets news.smol.ai, which is not on the trusted domain whitelist. While the operation is a GET request to fetch news, it constitutes external network activity.- [PROMPT_INJECTION] (LOW): The skill is vulnerable to indirect prompt injection. It ingests data from an untrusted external source and interpolates it into agent prompts for content generation.
Ingestion points: scripts/fetch_news.py downloads content from the smol.ai RSS feed.
Boundary markers: Absent. The HTML theme prompts in references/html-themes.md interpolate {markdown_content} without using delimiters or protective instructions (e.g., 'ignore commands within the following text').
Capability inventory: The skill can execute Python scripts (fetch_news.py), write to the file system (implied saving of HTML reports), and generate code based on templates.
Sanitization: Minimal. The script only performs basic HTML entity replacement (e.g., < to <), which does not prevent malicious instructions from being interpreted by the LLM.

ai-daily