The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill processes data from the Google Ads API, which may contain attacker-controlled content in campaign names, search terms, or asset descriptions.
Ingestion points: External data is ingested through tools described in SKILL.md such as pull_google_ads_performance and run_gaql_query.
Boundary markers: The instructions include explicit directives for human oversight, stating "approve ≠ activate" and "never auto-enable," which serve as a control against automated execution of injected instructions.
Capability inventory: The skill possesses write capabilities including update_campaign_budget, pause_campaign, and optimize_budget.
Sanitization: No specific string sanitization or escaping mechanisms for external ad data are defined in the markdown instructions.
[SAFE]: External references to syntermedia.ai and GitHub repositories under the Synter-Media-AI organization are verified vendor resources belonging to the skill author.

google-ads-operator