dbops-changeset

MEDIUM W011: Third-party content exposure detected (indirect prompt injection risk).

• Third-party content exposure detected (medium risk: 0.30). The skill ingests outsider-authored free text at runtime via the user’s natural-language description and follows it through intent classification and Liquibase YAML generation, with optional execution on “Accept & Commit”.

MEDIUM W012: Unverifiable external dependency detected (runtime URL that controls agent).

• Potentially malicious external URL detected (high risk: 0.80). The skill's reference.md explicitly instructs the agent at runtime to use webFetch to fetch Liquibase docs from https://docs.liquibase.com/change-types/.html (e.g., https://docs.liquibase.com/change-types/add-column.html), which is an external fetch used to control YAML generation and authoring decisions.