castai-debug-bundle
Pass
Audited by Gen Agent Trust Hub on Sep 13, 2026
Risk Level: SAFEDATA_EXFILTRATIONCOMMAND_EXECUTIONINDIRECT_PROMPT_INJECTION
Full Analysis
- [DATA_EXFILTRATION]: The skill performs network requests to the official service API at
api.cast.aito retrieve diagnostic metadata and cluster policies. It uses theCASTAI_API_KEYenvironment variable for authentication, transmitting it via HTTP headers as intended for this service interaction. - [COMMAND_EXECUTION]: The script executes a sequence of local system and Kubernetes management tools, including
kubectl,helm,tar, andjq, to aggregate cluster state and component logs into a timestamped archive. - [INDIRECT_PROMPT_INJECTION]:
- Ingestion points: Collects untrusted data from the cluster environment through
kubectl logs,kubectl get events, andhelm listoutput as defined inSKILL.md. - Boundary markers: Absent; collected log and event data are written directly to text files within the diagnostic bundle.
- Capability inventory: The skill possesses capabilities for shell command execution, file system operations, and outbound network access.
- Sanitization: None; the skill captures raw command output for troubleshooting purposes.
Audit Metadata