The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill's documentation (README.md and SKILL.md) instructs users to use npx -y mcp-server-google-analytics. This command downloads a package from the public NPM registry at runtime without verifying its integrity or pinning a specific version.
[REMOTE_CODE_EXECUTION]: By directing users to execute the mcp-server-google-analytics package via npx, the skill introduces a dependency on remote code that is not authored by the skill's creator. The security of the user's environment depends on the third-party package not being compromised.
[CREDENTIALS_UNSAFE]: The skill requires users to configure a GOOGLE_PRIVATE_KEY in their environment or MCP configuration file. This is a highly sensitive credential (RSA Private Key) that grants access to Google Cloud resources. The skill setup guides the user to expose this secret to the npx-delivered process.
[COMMAND_EXECUTION]: The skill requires the execution of shell commands (npx) to initialize its functionality, which is the primary vector for loading the required tools.
[INDIRECT_PROMPT_INJECTION]: The 'Conversion-Crater Detection' workflow (SKILL.md) instructs the agent to 'pull the article body HTML' and 'parse all CTA blocks'. Processing external, untrusted HTML content creates an attack surface where malicious instructions embedded in the webpage (e.g., in hidden tags or comments) could attempt to influence the agent's behavior during the audit. • Ingestion points: Google Analytics 4 API responses and article body HTML from local or remote web pages. • Boundary markers: None identified in the instructions to separate data from instructions. • Capability inventory: The skill uses tools to run reports and fetch analytics data (runReport, getPageViews, getActiveUsers, getEvents, getUserBehavior). The agent may also use other available system tools to fetch HTML. • Sanitization: No sanitization or validation steps are defined for the ingested HTML content.

google-analytics