The Agent Skills Directory

[COMMAND_EXECUTION]: Uses kubectl, helm, and tar to inspect cluster resources and package diagnostic data. These tools are used as intended for gathering system state and are explicitly listed in the allowed-tools configuration.
[DATA_EXFILTRATION]: Collects diagnostic data into a local archive for manual submission to support. The skill accesses the CAST AI API via curl to retrieve cluster status, which is a well-known service for this tool's context and is handled using a local environment variable.
[PROMPT_INJECTION]: The skill processes untrusted data from cluster logs and events. 1. Ingestion points: kubectl logs and kubectl get events in SKILL.md. 2. Boundary markers: None; collected data is stored in plain text files within an archive. 3. Capability inventory: kubectl, curl, helm, and tar. 4. Sanitization: Employs jq to filter API responses for non-sensitive fields and includes explicit instructions for user-led manual redaction of sensitive information before sharing.

castai-debug-bundle