The Agent Skills Directory

[REMOTE_CODE_EXECUTION]: The CI/CD examples in references/chaos-tools.md execute configuration directly from an external source via kubectl apply -f https://litmuschaos.github.io/litmus/litmus-operator-v2.14.0.yaml. This is a documented method for deploying the Litmus operator.- [COMMAND_EXECUTION]: The skill provides scripts that interact with high-privileged CLI tools, including aws ec2 terminate-instances for cloud infrastructure manipulation and kubectl delete for Kubernetes cluster management. These are used to simulate service disruptions as part of chaos experiments.- [PRIVILEGE_ESCALATION]: The references/infrastructure-chaos.md script uses sudo to append entries to /etc/hosts and to perform system-level operations like package installation with apt-get and resource testing with stress-ng.- [EXTERNAL_DOWNLOADS]: The skill references external binaries and manifests from well-known sources, including Litmus Chaos, Toxiproxy, and Gremlin APIs, across the reference documentation and implementation scripts.- [INDIRECT_PROMPT_INJECTION]: The skill possesses a powerful capability set (AWS, Kubernetes, and OS-level access) and ingests user-defined system data, creating an attack surface for instructions embedded in external logs or documentation.
Ingestion points: System descriptions and experiment parameters provided in SKILL.md and references/game-days.md.
Boundary markers: None identified in the provided templates.
Capability inventory: Subprocess calls for cloud provider APIs, container orchestration tools, and system resource managers.
Sanitization: No validation or sanitization of input data before its interpolation into command-line arguments or experiment manifests was found.

chaos-engineer