The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides numerous CLI commands for system manipulation, which are the primary intended function for chaos engineering.
Evidence includes network manipulation using tc qdisc (e.g., tc qdisc add dev eth0 root netem delay 200ms 50ms), infrastructure stress testing with stress-ng (e.g., stress-ng --cpu 4 --timeout 60s), and resource exhaustion using fallocate.
Kubernetes pod deletion via kubectl delete pod is also provided.
[INDIRECT_PROMPT_INJECTION]: The skill contains a pattern where it instructs the agent to ingest data from an external file, which constitutes an attack surface.
Ingestion points: The skill reads from .agents/qa-project-context.md to define the environment and architecture.
Boundary markers: No specific boundary markers or instructions to ignore embedded commands within the context file are mentioned.
Capability inventory: The skill enables significant system capabilities including network fault injection (tc, iptables), service disruption (kubectl delete), and resource exhaustion (stress-ng, fallocate).
Sanitization: No explicit sanitization or validation of the input from the context file is described.
This finding is categorized as a surface risk associated with the skill's intended operational workflow.

chaos-engineering