The Agent Skills Directory

[SAFE]: The skill documents the usage of Robusta, a legitimate and well-known tool for Kubernetes site reliability engineering. The content is instructional and aligns with the tool's official documentation.
[COMMAND_EXECUTION]: The skill provides numerous commands for cluster management and automation.
Includes usage of kubectl, helm, and robusta-cli for deployment and testing.
Documents node_bash_enricher and pod_bash_enricher actions, which allow SREs to automate the execution of bash commands on nodes or within pods for diagnostic purposes.
[EXTERNAL_DOWNLOADS]: The skill facilitates the download of necessary components from official and trusted sources.
Fetches the robusta CLI script from the official documentation domain docs.robusta.dev.
References the project's official Helm repository at robusta-charts.storage.googleapis.com.
[CREDENTIALS_UNSAFE]: Examples involving sensitive configuration (e.g., Slack bot tokens, PagerDuty keys, Jira API tokens) consistently use descriptive placeholders such as xoxb-your-token and your-api-token to prevent the exposure of real credentials.
[PROMPT_INJECTION]: While the skill mentions an AI-based investigation feature (HolmesGPT), it does not contain any patterns attempting to override agent instructions or bypass safety filters. The integration is presented as an optional feature of the Robusta platform.
Ingestion points: The platform processes alerts from Prometheus AlertManager and events from the Kubernetes API (identified in references/Triggers.md).
Capability inventory: The skill documents capabilities such as pod deletion (delete_pod) and command execution (node_bash_enricher and pod_bash_enricher), which are standard for automation tools but serve as an attack surface for indirect injection if alert data is not properly sanitized by the platform.
Sanitization: The documentation does not specify sanitization procedures, but notes a regex-based redaction feature in logs_enricher for sensitive data.

robusta-dev