The Agent Skills Directory

[SAFE]: No malicious patterns such as prompt injection, obfuscation, or unauthorized remote code execution were detected. The skill operates within the expected parameters of a professional marketing assistant.
[COMMAND_EXECUTION]: The skill utilizes external tools (referenced as ads.gaqlParallel and undoChange) to query performance data from the Google Ads API and manage ad deployments. These operations are performed in service of the skill's stated goals.
[DATA_EXPOSURE]: The skill accesses local files containing business context (business-context.json) and customer personas ({accountId}.json). This data ingestion is necessary for grounding the generated ad copy and is performed using standard file access patterns within the agent's data directory.
[PROMPT_INJECTION]: The instructions provide robust guidelines for ad copy generation and business logic (e.g., when to ignore Ad Strength metrics) without attempting to override the agent's core safety protocols or initial system instructions.

ads-copy