goal-engineering

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

• Third-party content exposure detected (medium risk: 0.30). The required runtime workflow reads a caller-supplied JSON file (task/method/context/etc.) from an external path and directly interpolates those free-text fields into the rendered goal string (scripts/render-goal.mjs → scripts/goal-core.mjs renderGoal), so outsider-provided text can be ingested without prior item selection.