The Agent Skills Directory

[COMMAND_EXECUTION]: The skill contains multiple Python scripts that utilize the subprocess module to execute system-level and administrative commands. This includes tools for disk space monitoring (df), service management (systemctl), and file cleanup (find).
[COMMAND_EXECUTION]: Extensive integration with Kubernetes is provided through the execution of kubectl commands. The skill demonstrates capabilities to delete pods, patch service specifications, and execute commands inside containers (e.g., kubectl exec for database operations or network manipulation).
[COMMAND_EXECUTION]: Chaos engineering scripts perform low-level system changes, such as using tc (traffic control) to inject network latency and iptables to create network partitions. These operations typically require elevated privileges.
[INDIRECT_PROMPT_INJECTION]: The skill's logic depends on data ingested from external sources, which could be manipulated to influence the agent's behavior.
Ingestion points: Data is collected from the output of system commands like df -h and kubectl get pods, as well as HTTP responses from curl health checks in the references/automation-toil.md file.
Boundary markers: No explicit boundary markers or instruction-ignoring delimiters are used when processing this external data.
Capability inventory: The skill has broad execution capabilities through subprocess.run, including network access and system-level configuration changes.
Sanitization: While the scripts avoid shell=True to prevent direct shell injection, they rely on standard string splitting and basic parsing which may be susceptible to logic manipulation if the command outputs are attacker-controlled.

sre-engineer