The Agent Skills Directory

[SAFE]: The skill consists of natural language instructions for troubleshooting infrastructure incidents. It does not include executable scripts, hardcoded credentials, or remote code downloads.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to analyze external data sources such as logs and metrics. Ingestion points: query-logs, query-metrics, query-apm-spans. Boundary markers: absent. Capability inventory: read-only observability tools. Sanitization: absent. While this creates a potential surface for indirect injection, the risk is minimal as the tools are limited to data retrieval and the behavior is inherent to the skill's primary diagnostic purpose.

investigating-metric-anomalies