The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill installs the google-ads-mcp package directly from the author's GitHub repository via pip install git+https://github.com/itallstartedwithaidea/google-ads-mcp.git.
[COMMAND_EXECUTION]: The server is executed locally using the python -m ads_mcp.server command as part of the MCP configuration.
[PROMPT_INJECTION]: The skill exhibits an indirect prompt injection surface as it ingests untrusted data from the Google Ads API and has the capability to perform account modifications.
Ingestion points: Untrusted data enters the agent context through tools like get_search_terms, get_ad_performance, and execute_gaql which fetch content from Google Ads.
Boundary markers: The instructions do not specify any delimiters or safety warnings to distinguish external API data from system instructions.
Capability inventory: The skill includes several high-privilege write tools such as update_campaign_budget, create_campaign, and generic_mutate which can modify account state.
Sanitization: There is no evidence of data sanitization or validation for content retrieved from the API before it is processed by the agent.

google-ads-mcp