The Agent Skills Directory

[SAFE]: The skill implements protective measures against image beaconing in its visual review templates. The assets/creative-review-template.html file includes logic to block remote and protocol-relative URLs, ensuring that the report remains private and does not leak metadata to external servers.
[SAFE]: The documentation and templates provide clear instructions on securing data rendering, specifically recommending character escaping for HTML tags to prevent cross-site scripting (XSS) or data breakout when generating dynamic reports.
[INDIRECT_PROMPT_INJECTION]: The skill is designed to ingest and process external content such as customer reviews and ad platform comments. It addresses the inherent risk of indirect prompt injection by mandating a 'grounded input' strategy, where the agent is instructed to cite specific source identifiers for all claims and copy, thereby maintaining a strict boundary between user-provided data and agent instructions.

ad-creative