The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill is designed to interact with and process data from external Google Tag Manager (GTM) accounts, which acts as an untrusted input surface.
Ingestion points: Data enters the agent's context through several GTM-specific tools, such as mcp__cogny__google_tag_manager__tool_get_tag, mcp__cogny__google_tag_manager__tool_get_trigger, and mcp__cogny__google_tag_manager__tool_get_variable, which fetch live configuration details from the user's environment.
Boundary markers: The instructions do not specify any delimiters or safety warnings to distinguish between the skill's instructions and the external data retrieved from the GTM API.
Capability inventory: The agent possesses a broad range of capabilities including Bash command execution, file Write operations, and network requests via WebFetch. These could be targeted for abuse if a GTM container contains malicious metadata or descriptions.
Sanitization: The skill lacks explicit sanitization, filtering, or validation steps for the content fetched via the GTM MCP tools before the agent processes or acts upon that content.

gtm-setup