The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses the Bash tool, restricted to curl, to interact with VictoriaMetrics, VictoriaLogs, and VictoriaTraces API endpoints. These operations are performed using environment variables for base URLs and authentication headers, which is the recommended approach for managing infrastructure credentials and endpoints safely.
[PROMPT_INJECTION]: The skill processes external data such as application logs, traces, and alerts, which constitutes an indirect prompt injection attack surface. 1. Ingestion points: Data is fetched from VictoriaMetrics, VictoriaLogs, and VictoriaTraces API endpoints into the agent's context (see agents/*.md files). 2. Boundary markers: None identified in the prompt templates; data is processed as raw output. 3. Capability inventory: Bash(curl:*), Agent (subagent dispatch), and Read (file access) across all investigation phases. 4. Sanitization: The skill does not implement explicit sanitization of the content retrieved from observability backends. Note: This vulnerability surface is inherent to the skill's purpose as a diagnostic tool and does not indicate malicious intent in the skill itself.

investigating-with-observability