mirrord-db-branching
Pass
Audited by Gen Agent Trust Hub on Oct 22, 2026
Risk Level: SAFE
Full Analysis
- [DATA_EXPOSURE_AND_EXFILTRATION]: The skill incorporates clear security boundaries regarding sensitive information. It explicitly directs the agent to never hardcode credentials, passwords, or literal connection strings. Instead, it promotes the use of environment variables, Kubernetes Secrets, and cloud-native secret managers (AWS Secrets Manager and Google Secret Manager). It also cautions against asking users for passwords directly.
- [INDIRECT_PROMPT_INJECTION]: While the skill ingests and processes user-provided configuration data, it includes defensive instructions. The agent is told to treat all user-provided values as untrusted data and is specifically forbidden from executing shell commands or SQL derived from those values, effectively neutralizing the risk of command injection via configuration.
- [COMMAND_EXECUTION]: The skill employs local command execution (
mirrord verify-config) for the purpose of validating generated configurations. This is an intended and safe use of the mirrord CLI tool within its operational context. - [EXTERNAL_DOWNLOADS]: The skill references external resources such as official documentation from the vendor's domain and well-known container images for database migrations (Flyway, Liquibase). These are standard technology resources and do not pose a security risk in the context of this skill.
- [DYNAMIC_EXECUTION]: The skill provides instructions for configuring container-based migrations and generic branches, which involves defining images and entrypoints. The documentation correctly notes that these are cluster-admin level configurations and provides guidance on administrative gating and allowed image policies to maintain security.
Audit Metadata