chaos-engineer

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (high risk: 0.90). The skill's required CI/CD workflow and runbook examples explicitly fetch and execute remote, public manifests and APIs (for example the GitHub Actions step "kubectl apply -f https://litmuschaos.github.io/..." in references/chaos-tools.md and the GremlinClient calls to https://api.gremlin.com), meaning untrusted third‑party content would be ingested and acted on and could therefore influence tool actions.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.90). The GitHub Actions workflow includes a runtime step that runs "kubectl apply -f https://litmuschaos.github.io/litmus/litmus-operator-v2.14.0.yaml", which fetches remote YAML and executes it on the cluster (a required dependency for the subsequent chaos experiments), so this external URL executes remote code at runtime.

MEDIUM W013: Attempt to modify system services in skill instructions.

• Attempt to modify system services in skill instructions detected (high risk: 0.90). The skill explicitly directs the agent to design and execute failure-injection experiments and produce implementation code/manifests that intentionally modify or disrupt system and infrastructure state (processes, services, network rules, Kubernetes pods, etc.), actions which commonly require elevated privileges or edits to system-level config and therefore push the agent toward compromising the host state.