The Agent Skills Directory

[COMMAND_EXECUTION]: The skill utilizes several standard command-line tools, including kubectl, flux, and gh, to manage and inspect deployment states. It references sensitive Kubernetes configuration files in the ~/.kube/ directory (e.g., integration.yaml, live.yaml) to switch cluster contexts.
[PROMPT_INJECTION]: The skill contains an attack surface for indirect prompt injection by instructing the agent to ingest data from external, potentially untrusted sources.
Ingestion points: The agent is directed to retrieve content from GitHub Action logs using gh run view --log and to inspect Kubernetes resource metadata using kubectl describe or kubectl get -o yaml.
Boundary markers: No specific boundary markers or instructions are provided to the agent to treat external log content or resource descriptions as data rather than instructions.
Capability inventory: The skill grants the agent the ability to execute system commands, modify cluster states via flux and kubectl, and interact with repository workflows.
Sanitization: There is no logic provided to sanitize or validate the content retrieved from GitHub logs or cluster objects before it is processed by the agent.

promotion-pipeline