The Agent Skills Directory

[SAFE]: The skill follows best practices for summarizing external content by requiring the agent to fetch the URL rather than guessing content. It uses established tools for data retrieval.
[PROMPT_INJECTION]: The skill processes untrusted data from external URLs, which presents a surface for indirect prompt injection. However, it mitigates this risk by enforcing a 'Quote-Grounding Technique' where key passages must be extracted before a summary is generated, allowing for verification of claims against source text.
[SAFE]: The skill references local configuration and template files using relative paths. This is standard behavior for skill-based architectures to maintain structured outputs.

url-summarization