The Agent Skills Directory

[SAFE]: The skill defines a structured methodology for gathering information from verified and well-known news organizations, research blogs, and academic repositories. All listed external sources in references/sources.md are reputable and align with the skill's objective of providing high-quality AI updates.
[PROMPT_INJECTION]: The skill interacts with untrusted data from public community forums (e.g., Reddit, Hacker News) and newsletters using the web_fetch tool. This presents a surface for indirect prompt injection, where an attacker could potentially influence the agent's output by placing malicious instructions on these sites. However, the risk is evaluated as minimal because the skill's capabilities are limited to information retrieval and summarization, with no access to sensitive system tools, credentials, or file-writing operations. The analysis identified the following evidence chain: ingestion occurs via web_fetch in SKILL.md; no explicit boundary markers or 'ignore' instructions are used; capabilities are restricted to search/fetch operations; and no content sanitization is implemented.

ai-news-collector