The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill implements a /google-ads landing-review command which uses external tools to fetch and process content from landing page URLs.
Ingestion points: Untrusted data enters the agent context via the browser fetch results when analyzing landing pages as described in the routing logic for the landing-review command.
Boundary markers: There are no explicit instructions or delimiters defined to separate the fetched content from the agent's internal instructions or to warn the agent to ignore embedded commands within the landing page HTML/text.
Capability inventory: The skill utilizes the google-ads-mcp tool to perform GAQL searches and has the capability to write draft files to the workspace/ads/ directory. The apply command provides the capability to execute account changes (pausing keywords, adding negatives) via the Google Ads API.
Sanitization: The instructions do not specify any validation or sanitization protocols for the data retrieved from external URLs before it is processed for intent mapping and diagnostic reviews.

google-ads