The Agent Skills Directory

[INDIRECT_PROMPT_INJECTION]: The skill ingests data from external cluster resources, such as pod logs, events, and resource descriptions, which could potentially be manipulated by a malicious actor with cluster access to influence agent behavior.
Ingestion points: SKILL.md (logs/describe commands), scripts/cluster-health-check.sh (health assessment), and scripts/pre-upgrade-check.sh (version and status checks).
Boundary markers: Absent. Cluster data is presented directly to the agent context without delimiters.
Capability inventory: High-privilege cluster management capabilities including manifest application, node draining, and cloud CLI usage (AWS, Azure, GCP).
Sanitization: None. While scripts use jq for parsing structured data, raw text logs and event messages are displayed as-is.
[COMMAND_EXECUTION]: The skill relies on executing shell commands via kubectl, oc, and various cloud CLIs. This is the core intended functionality of the skill for cluster management and is conducted using standard, transparent bash scripts included in the package.

kubernetes