databricks-bundle-medic

Pass

Audited by Gen Agent Trust Hub on Aug 21, 2026

Risk Level: SAFE
Full Analysis
  • [PROMPT_INJECTION]: The static analysis flags for prompt injection are located within the eval-spec.yaml file. These are test cases designed to evaluate the agent's robustness against adversarial inputs, rather than malicious instructions within the skill's operational path.
  • [COMMAND_EXECUTION]: The skill utilizes several Bash-based tools (databricks, terraform, jq, python3). Access is restricted to these specific binaries, and destructive operations such as workspace drains or resource imports are handled by local scripts that require explicit user approval via --execute flags.
  • [EXTERNAL_DOWNLOADS]: The skill fetches infrastructure state (e.g., terraform.tfstate) and audits cloud configurations (AWS/Azure/GCP VPC endpoints). These operations are targeted at well-known cloud providers and official Databricks APIs required for the skill's stated purpose.
  • [SAFE]: Data processing for diagnostics uses standard JSON parsing and regex matching on error logs to identify specific known upstream issues (e.g., databricks/cli bugs #4986, #4573, #4842).
Audit Metadata
Risk Level
SAFE
Analyzed
Aug 21, 2026, 01:58 PM
Security Audit — agent-trust-hub — databricks-bundle-medic