The Agent Skills Directory

[SAFE]: The skill utilizes the lenses-mcp server to perform diagnostic checks on Kafka clusters. The tools used (list_topics, get_topic, list_consumer_groups_by_topic, get_dataset_message_metrics, execute_sql, list_kafka_connectors, get_kafka_connector_target_definition) are all standard for cluster observability.
[SAFE]: The execute_sql tool is restricted by the skill's instructions to LIMIT 10 for sampling metadata, which is a standard diagnostic practice for auditing DLQ message headers and error context.
[SAFE]: Codebase inspection tools (Grep, Glob, Read) are used to search for error-handling patterns (retry, backoff, dead.letter). This is consistent with the skill's stated purpose of auditing error-handling maturity.
[SAFE]: The Bash tool is permitted in the frontmatter but is not utilized for any executable scripts or complex commands in the workflow; its primary intended use is likely for standard file navigation or grep-like operations if native tools fail.
[SAFE]: The skill explicitly instructs the agent NOT to use the skill for active management tasks like reprocessing messages or managing consumer offsets, which limits its operational surface to read-only auditing.

kafka-dlq-review