The Agent Skills Directory

[SAFE]: The skill content consists purely of instructions and benchmarks for ad campaign optimization. There are no instances of code execution, network exfiltration, or unauthorized file access.
[PROMPT_INJECTION]: The skill processes untrusted landing page URLs, creating an indirect prompt injection surface. The ingestion point is the campaign URL collection step; there are no boundary markers or sanitization mentioned in the instructions. However, the capability inventory is limited to scoring and report generation, and the instruction set is focused on objective evaluation of quality metrics, which represents a low-risk surface.
[SAFE]: File access to 'ads/references/' is limited to reading internal benchmarks and tracking configurations necessary for the stated purpose of the skill.

ads-landing