The Agent Skills Directory

[COMMAND_EXECUTION]: The skill facilitates the execution of numerous shell commands and local scripts including make craft-up, make craft-migrate, and deployment/helm/dev/craft-up.sh. These commands perform complex operations such as building container images, installing Helm releases, and managing Kubernetes clusters.
[PRIVILEGE_ESCALATION]: The instructions note that the telepresence networking daemon may require a sudo password to establish connections. While the skill warns that automated tasks cannot provide this password, it establishes a workflow involving elevated privileges.
[INDIRECT_PROMPT_INJECTION]: The skill instructs the agent to inspect Kubernetes pod logs and events to debug failures. This creates a surface for indirect prompt injection where malicious content within container output could attempt to influence the agent's behavior.
Ingestion points: kubectl logs, kubectl get pods, and pod events (SKILL.md).
Boundary markers: No specific delimiters or warnings for the agent to ignore instructions embedded in the logs are provided.
Capability inventory: Extensive command execution capabilities including make, kubectl, kind, helm, and python (SKILL.md).
Sanitization: No evidence of sanitization or filtering of log/event content before processing.
[CREDENTIALS_UNSAFE]: The skill interacts with sensitive environment files such as .vscode/.env.k8s and .vscode/.env.web. It explicitly mentions that these files should be kept private as they contain secrets and overrides used during the launch process.

onyx-craft-dev