The Agent Skills Directory

[EXTERNAL_DOWNLOADS]: The skill provides instructions for adding the official Grafana Helm repository (grafana.github.io) and utilizing standard OpenTelemetry libraries for Java, Python, Go, and Node.js. These references point to established, legitimate software sources within the observability ecosystem.\n- [COMMAND_EXECUTION]: Includes standard operational commands for helm, kubectl, and the Azure CLI (az) required for deployment, monitoring, and troubleshooting of a Loki cluster. These commands are appropriate for the documentation's technical purpose.\n- [CREDENTIALS_UNSAFE]: The documentation follows security best practices by using appropriate placeholders (e.g., <access-key>, <account-key>) for sensitive credentials and recommends the use of managed identities (e.g., Azure Managed Identity) over long-lived keys for authentication.\n- [PROMPT_INJECTION]: The skill outlines methods for processing log data from external sources. While there are no direct injection vulnerabilities, the processing of untrusted logs presents a surface for indirect prompt injection.\n
Ingestion points: System and application logs enter the system through the Loki Push API or OpenTelemetry Collector as described in SKILL.md and references/opentelemetry.md.\n
Boundary markers: The skill does not define specific delimiters for isolating log content from instructions in its examples.\n
Capability inventory: The skill demonstrates capabilities for querying, extracting, and reformatting log data using LogQL across SKILL.md and references/logql.md.\n
Sanitization: No explicit sanitization or filtering of log content for AI instruction detection is described.

loki