The Agent Skills Directory

[COMMAND_EXECUTION]: The skill provides playbooks that use kubectl and curl for service discovery, remediation, and environment modification. Examples include kubectl set env, kubectl rollout undo, and kubectl patch configmap. While standard for SRE operations, these commands require high-privilege access to production infrastructure.
[INDIRECT_PROMPT_INJECTION]: The skill possesses a surface for indirect prompt injection through its data processing workflows.
Ingestion points: The agent is instructed to read and analyze potentially untrusted data from kubectl logs, external provider status pages, and Prometheus metrics or alerts.
Boundary markers: The instructions include manual verification steps and warnings about PII exposure, but they do not define technical boundary markers or specific "ignore embedded instructions" delimiters for the logs being processed.
Capability inventory: The skill has high-privilege write access to the environment, including the ability to modify environment variables, rollback deployments, and patch system configurations via kubectl.
Sanitization: There are no explicit sanitization or filtering steps defined for the external log data or provider responses before they are used to determine remediation actions.

ai-sre-incident-response