The Agent Skills Directory

[Prompt Injection] (SAFE): No attempts to bypass safety filters or override system instructions were detected. The markdown follows standard instructional formatting.
[Data Exposure & Exfiltration] (SAFE): The skill references local data files (ads/references/) for benchmarking but does not access sensitive system paths (like SSH keys or credentials) or perform network exfiltration.
[Remote Code Execution] (SAFE): No commands for downloading external scripts or executing arbitrary code (e.g., curl | bash or eval) are present.
[Privilege Escalation] (SAFE): The skill does not use sudo, chmod, or any other mechanism to escalate user privileges.
[Indirect Prompt Injection] (SAFE): While the skill ingests external Google Ads export data, it lacks dangerous capabilities (such as shell execution or file system modification) that would allow such data to exploit the agent. The process is purely analytical.

ads-youtube