teamharness-task-execution

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

• Third-party content exposure detected (high risk: 0.70). The skill reads shared/tasks/{task-id}/spec.md and shared/tasks/{task-id}/meta.json at runtime; if those files are authored by the Leader/outsider rather than the operating user, their free-text contents can be ingested into the agent’s LLM context via the task-directory file reads.