The Agent Skills Directory

[COMMAND_EXECUTION]: The skill uses sudo in references/infrastructure-chaos.md to install the stress-ng package and modify the system /etc/hosts file for DNS failure simulation. These operations involve privilege escalation.
[COMMAND_EXECUTION]: Provides numerous scripts and templates in references/chaos-tools.md, references/game-days.md, and references/kubernetes-chaos.md that use the aws CLI and kubectl to terminate EC2 instances, reboot RDS databases, and delete Kubernetes pods. While functional for chaos engineering, these are destructive operations.
[EXTERNAL_DOWNLOADS]: The skill downloads the Litmus Chaos operator manifest directly from https://litmuschaos.github.io/litmus/litmus-operator-v2.14.0.yaml and applies it using kubectl. It also uses brew to install toxiproxy and apt-get for stress-ng.
[PROMPT_INJECTION]: The skill contains an indirect prompt injection surface (Category 8) where it ingests and processes external data such as application logs, Prometheus metric queries, and Kubernetes event manifests.
Ingestion points: Log tailing (kubectl logs), Prometheus API responses (requests.get), and resource status checks (kubectl get chaosresult).
Boundary markers: None identified in the provided templates to isolate processed data from the agent's core instructions.
Capability inventory: The skill has extensive access to cloud APIs (AWS), container orchestration (Kubernetes), and local shell execution.
Sanitization: The scripts do not explicitly demonstrate sanitization of the metrics or logs before they are presented to the agent for analysis.

chaos-engineer