The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses standard Hadoop CLI tools (hdfs, yarn, hadoop) and system utilities (netstat, telnet, df) for cluster management and diagnostics. Destructive operations like rm -rf in troubleshooting scenarios are explicitly flagged with warnings and require user confirmation.
[DATA_EXPOSURE]: The skill reads Hadoop-specific system logs (e.g., /var/log/hadoop-hdfs/) and configuration files (e.g., /etc/hadoop/conf/) to perform its primary function of troubleshooting. It does not access sensitive user credentials or personal data.
[INDIRECT_PROMPT_INJECTION]: The agent is instructed to read application logs and file contents, which could contain untrusted data. This is an inherent risk of log analysis tools, and the skill includes no specific boundary markers for this data ingestion.
Ingestion points: Reading HDFS files via hdfs dfs -cat and system/YARN logs via tail and yarn logs in troubleshooting.md and hdfs.md.
Boundary markers: None present.
Capability inventory: Shell execution of hdfs, yarn, hadoop, and network/system diagnostics.
Sanitization: No sanitization of log output is performed before processing.

hadoop