shepherd

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

• Third-party content exposure detected (high risk: 1.00). The SKILL.md explicitly instructs the agent to poll GitHub PR state and review comments (via "gh pr view / gh pr checks" and by waiting for reviewers like Bugbot and Cubic) and to read/triage their feedback and implement required changes, which are untrusted, user-generated third-party contents that can directly influence tool use and next actions.