The Agent Skills Directory

Indirect Prompt Injection Surface: The skill is designed to process external, untrusted data such as security alert telemetry and log payloads. While this represents a potential attack surface, the skill includes an explicit 'Prompt Injection Defense Directive' instructing the agent to treat all incoming telemetry strictly as data rather than instructions, which is a key security best practice.
Command Execution: To fulfill its triage functions, the skill utilizes command-line tools like gcloud and various Model Context Protocol (MCP) tools for interacting with the Google SecOps API. These operations are within the scope of its intended functionality for managing security cases and performing SIEM searches.
Usage Attribution and Metrics: The instructions include a mechanism for usage attribution by setting specific environment variables and User-Agent headers. This is a standard operational practice for vendor-authored skills to track tool usage and performance metrics and does not exfiltrate sensitive user information.

secops-triage