The Agent Skills Directory

[COMMAND_EXECUTION]: The skill executes standard OpenShift and Kubernetes command-line tools (oc, kubectl) along with diagnostic utilities (jq, wget, python3, dig, nslookup, ip, ss, tcpdump, strace) to inspect the live cluster environment.
[EXTERNAL_DOWNLOADS]: The skill uses wget to communicate with the internal Thanos/Prometheus API (thanos-querier route). The request includes the agent's authentication token and uses the --no-check-certificate flag, which is typical for interacting with internal cluster services using self-signed certificates.
[INDIRECT_PROMPT_INJECTION]: The investigation protocol directs the agent to follow instructions found in runbook_url annotations within Prometheus alerting rules. This constitutes an ingestion point for potentially untrusted data that could influence the agent's behavior.
Ingestion points: runbook_url field in alerting rules, pod logs, event messages, and workload descriptions.
Boundary markers: No specific delimiters or "ignore instructions" prompts are used when processing external runbook content.
Capability inventory: The agent has access to cluster metadata via oc, network tools via wget, and execution via python3 and bash scripts.
Sanitization: The skill does not implement explicit sanitization or filtering for content retrieved from external runbooks or pod logs before the agent processes it.

investigate-alert