The Agent Skills Directory

COMMAND_EXECUTION (HIGH): The skill relies on high-privilege kubectl and helm commands for infrastructure management. A PreToolUse hook in SKILL.md is configured to execute .claude/hooks/verify-kubectl-context.sh, but this script is not included in the skill's file list. This represents an unverifiable execution of logic with the agent's full system access.
PROMPT_INJECTION (HIGH): The skill exposes a large Indirect Prompt Injection surface (Category 8). It provides runbooks and templates for consuming and analyzing Kafka messages and logs (e.g., debugging-runbooks.md, Makefile, and producer-consumer.py). Since the agent has kubectl exec and cluster modification capabilities, malicious instructions embedded in Kafka topics could be processed by the agent, potentially triggering unauthorized cluster operations without explicit user consent.
EXTERNAL_DOWNLOADS (MEDIUM): The skill configures external downloads from strimzi.io (Helm charts) and maven.org (Debezium plugins). These sources are not included in the predefined list of Trusted External Sources and thus represent unverifiable third-party dependencies.
INSECURE_CONFIGURATION (MEDIUM): The default cluster manifest in manifests/kafka-cluster.yaml exposes an external Kafka listener via nodeport without TLS encryption or authentication. This default configuration is highly insecure for production or non-isolated environments.

kafka